Issues - GC5325SEK

Hello,

The GC5325 DPD engine can apply CFR and DPD to a composite signal.  The static GSM signals supplied with the GC5325SEK, could be used for a low power test, and then with a higher power power-amplifier.

The GC5322 is used by some customers with custom software in a DPD application for GSM.  The GC5322 and GC5325 share the same CFR and DPD engine.  The GC5325SEK does not come with this special software.   If you have a non-hopping GSM application, (which is similar to the supplied 6 carrier test case), the suggestion is to run a test case using the supplied waveform.   Ken Chan kenchan@ti.com in the Solutions group may have a GSM report, there is not a published report.  I will pass this email on to Ken as well. 

Related to multicarrier GSM and hopping frequencies, the newer product GC5330 has a hopping generator and different method to apply CFR for this case.  The GC5330

EVM software for GSM is still being developed. 

2.  The EVM system uses a clock which is divided for several devices.  The DAC5682 uses the high speed clock 614.4 to 737.28.  Since the DAC interpolates by 4, the IQ rate from the GC5325 is set for 153.6 to 184.32.  The GC5325 is programmed for a specific interpolation after DPD of 1.5.  The DPD clock is 2x the DPD rate of 204.8 to 245.76 (for a DPD IQ rate of 102.4 to 122.88).  The Farrow Resampler resamples the Baseband composite input at a selected rate to the DPD rate.  Another clock at 2x the Baseband rate is sent to the TSW3100 for the Baseband data output. 

In addition the analog IF filters are based on the DAC Fs/4 mix of the 614.4 to 737.28Mhz clock.   As part of  your customization process, the IFs can be modified, if the Fs/8 mix of the DAC is used.  The DAC IF must be the nyquist sampled feedback IF.  Typically the EVM is build with a more narrow set of IF constraints to help filter the IF signal. 

3.  The range of 6 to 7 dbm for LO drive was based on previous use of the GC5322 EVM.  As you have analyzed the RF mixer and its drive level, the highest allowable

LO drive for a +4dbm input would seem to be -.69dbm.  I suggest that you try a range of -7 to 1dbm and select the value that gives the best performance.    The Cal Feedback tries to balance the feedback input level to the reference level.   If you get a manual capture buffer from the DPD reference to the Feedback after equalization the power level can be matched to within 2 db manually.  If the attenuator or gain is still appropriate, you can Cal Feedback to automatically adjust this to a +/-.25db error.  A failure in Feedback calibration usually occurs if the signal strength is not within the range of the feedback attenuator to adjust.  (another case is that we are adjusting the wrong antenna stream). 

Regards,

Radio Joe

Hi Joe,

Thanks a lot for the information.

We'd like to evaluate the GC5325SEK for a non-hopping GSM application. I hope the GSM test files are sufficient for this purpose. Do we have to use a external clock while evaluating for GSM since the default VCXO (737.28 MHz) is not a multiple of the GSM symbol rate of 270.833 KHz ? I'll contact Ken Chan and get the GSM report as you suggested.

You have mentioned in your reply that the GC5325 has been programmed for a specific interpolation of 1.5x after DPD. I'd like to know whether the interpolation factors set for the Farrow resampler and the Bulk Upconverter are 2x and 1.5x respectively. Besides I'd also like to know whether both of them can be configured using the GUI for the GC5325SEK.

I know it is possible to correct for dc offset using the GUI. Is it possible to correct quadrature imbalances as well ?

I would also like to get more information on the functions and configuration of some of the modules in the GC5325SEK. I could not get detailed information on these omodules from the documents I have come across. They are given below.

1. Circular Limiter

2. Feedback NL correction and Feedback Equalizer

3. Transmit Equalizer

Regards,

Prem Krishnan

Hi Prem,

The MCGSM static files are sampled at 26Msps.  This can be entered in the GUI in the Input Rate (MSPS) section instead of the default 61.44.  This will tell the GUI to resample the input signal to match the existing frequency plan and GC configuration.  This is all thats needed to use the MCGSM files included with the GUI software.  Everything else is the same process as described in the users guide.

The GUI does have basic DC offset controls - you can access these by clicking on the Offset Cal button in the DAC5682 section of the GUI.  For DAC QMC and analog Gain you will have to do a register write via the R/W button in the same DAC section.  This is not the most optimal way to do this as it is quite tedious to write values to registers.

For the static MCGSM results did you have a particular carrier configuration.  I remember we did testing with 6c evenly spaced within 5M, 10M, 15M BW.  Let me know if you have a particular preference and I will try to find that data.

Regards,

Ken

Hello,

Please consider starting a GC5330 thread, I will try to answer the question directly.

The GC5330 DPD has several components, a generic runtime for the GC5330 EVM, with a Transmit chain static frequency Tx, and a repeater EVM example.

There is a customer specific development program for GSM (static and hopping frequency).  If you would like more details for a customer specific Hopping GSM, these are not public forum topics.

Please contact your local Texas Instruments Field Application Engineer or Distributor for Texas Instruments.  You cal also contact Jay Shu (jshu@ti.com) the embedded

DSP lead engineer in the Radio Products group.

Thank you for your interest in the GC5330 product.

Regards,

Radio Joe

Hello,

The GC5325 input is considered to be from a Digital Up Converter - Interpolation, FIltering, Mixing, summed carriers, and gain adjustment.   In the EVM the TSW3100 provides a static pattern that is replayed.  Given the nature of the static pattern, if you build a waveform then resample and mix, not only does the length of the TSW3100 need to be the

TSW3100 pattern length = Baseband IQ pattern length * interpolation * 2

The mixing frequencies selected for each carrier must be a 2*pi phase multiple over the number of samples. 

As an example we want to transmit LTE10 (BB rate 15.36e6) to the GC5325 input. 

The GC5325 input is at a rate of 61.44Msps, the TSW3100 memory system must run at 122.88Msps to transfer the interleaved complex data.

The TSW3100 pattern length for 10ms data period would be 122.88e6 * 10e-3 = 1.2288e6 memory elements, .6144e6 complex samples.  You would create a two column vector output file with .6144e6 samples.  In practice, I usually take 2 Tx frames, interpolate, mix, sum, and scale, and then crop the file to exactly one output frame.

You would need to adjust the tuning frequency in the software DUC so that the phase wraps at the sample boundary

% ************************************************************************************

% CoherentFreq

% adjusts frequency to closest frequency that has a complete cycle within numtestpts

% ************************************************************************************

function AdjFreq = CoherentFreq( InFreq, Fs, numtestpts)

AdjFreq = (InFreq/abs(InFreq))*max(primes(round(abs(InFreq)/Fs*numtestpts)))/numtestpts*Fs; %Adjusted frequency, works also for negative frequencies

return ;

Regards,

Radio Joe

Hello there Radio-Joe,

I was off line for a good while recovering from my last 5 years non-stop effort you know where.  I may start a 5330 thread later.

For now, I know you have been a straight shooter with me all the way.  This customer specific GSM/MCGSM static and Hopping is a big Q from me.  If such a thing will one day exist from TI, it will be profitable fo all potential customers, said TI brass.  Until then, you should advise all your customers who has the key formula for such.  And make no mistake, erasing my posts, WILL NOT ERASE THIS FACT, said me.

Regards,

DSP_Specialist.

Hello,

Several customers have requested either a GSM static or GSM hopping demonstration, on the GC5325 platform where this thread started.  The current GC532x EVM can be obtained through your local TI HSP FAE, or by contacting Ken Chan or Jim Seton.  The current GC5325 EVM software does not support Hopping GSM.

The GC5325 SEK is a TI first generation DPD, and feedback processor.  The GC5325 has a second generation CFR processor also.  In most cases, the bandwidth

is limited to the 1.4x oversampling requirement for CFR, and the 130Mhz BB clock.  The GC5325SEK does come with a sample static GSM waveform, and if you import scaled I and Q data (signed decimal two columns) other static waveforms can have CFR and IMD correction.

When the GC5325 is implemented the DUC interpolation, filtering, mixing, summing, and scaling must be performed in the customer's logic. 

Please contact Ken Chan, if you want to discuss third party services.  It is the forum policy to allow this with approved 3rd party vendors.  Previous posts offering services without this approval were removed.

Regards,

Radio Joe

Hello,

The GC533x has specific functions tailored for signals with frequency hopping, and in some cases power level changes.

The DDUC (Digital Up Down Converter) has a hopping sequence table, that allows the user to program a set of 64 different frequencies, and indexes to those 64 frequencies.  The timing generator can change the index to the frequency tables with a hardware counter for the frequency update.    The sequence table changes the phase update word for the Numerically Controlled Oscillators.

The CFR (Crest Factor Reduction) block has a dynamic mode that allows the power level or frequency shifted carriers to be clipped with a modified copy of the signal.  This removes the requirement for updating the cancellation coefficients related to frequency changes.  The CFR block has multiple clipping thresholds and levels, currently the selection of the threshold and levels is done through register selection.

The Transmit equalizer and Non Linear DPD (Digital PreDistortion) updates have two choices, in the initial design, the DPD solution could operate with a combined update over the operating band.  This method has several db lower performance than the per frequency / power update approach.  If the customer wants to update the Transmit Equalizer and Non Linear DPD on a per time slot basis, custom DSP software is needed for this.  The frequency / power DPD update software has special licensing, please contact Jay Shu. jshu@ti.com.

If the feedback equalizer requires changing, for frequency hops, the filter is performed in the DSP (Digital Signal Processor). 

Regards,

Radio Joe

Hello,

Please see the post starting with "

Hello,

The GC533x has specific functions tailored for signals with frequency hopping, and in some cases power level changes."

Regards,

Radio Joe

Hello,

The GC5330 EVM is a demonstration tool for the building and playing a static Tx input signal (TSW3100), the main board that has the GC533x, 6748DSP, Ethernet Host interface, and depending on the customer configuration, the internal Data Converter and RF components, or the external Data Converter and RF components.

With a specialized TSW3100 the DDC receive path can also be captured for analysis. 

The standard adaption model is supported with a standard software license.  The frequency hopping GSM adaption model is supported with a modified EVM, with a special software license and hardware configuration.   Your requirements should be discussed with your local TI FAE and/or the DSP software lead engineer Jay Shu jshu@ti.com.

The customer is free to build an accelerated EMIF bus, to store and forward programming of the GC533x.  This would require the DSP to interface to another device, which stores and send at specific times the DMA sets for the next adaption solution. 

Regards,

Radio Joe 

Hello,

The CFR for the GC5325 (this thread) uses the standard GC532x CFR, which has a main memory  and shadow memory to use the frequency-dependent, and pre Farrow processed cancellation coefficients.  In a frequency or power hopping system, the coefficients can be changed for the GC532x by updating the shadow memory coefficients, using a software API.  This usually requires a hardware synchronization for the update, using the SyncB or SyncC signal, as the SyncA signal is used for the Tx synchronization. 

The CFR for the GC533x uses either the main memory and shadow memory for frequency dependent coefficients, or in dynamic mode, the coefficients are convolved with the signal.   In a frequency or power hopping application, the single set of dynamic coefficients is utilized.  The PostCFR capture buffer can be used to tune the preCFR and postCFR gain to control the amount of CFR applied. 

The GC533x family CFR application note, is available under an NDA for the GC5330.  Please contact you local TI Field Application Engineer or Kyle Li (kli@ti.com).  Once you have an NDA we can provide the application note, and illustrate the CFR programming code that is needed for your system.  In most cases, the CFR programming code is part of a configuration for the TI EVM.  The software for controlling CFR related to the PreCFR gain, PostCFR gain, Capture Buffer Power meter API, Capture Buffer API, are in the Software User Guide section for APIs.  Please contact Jay Shu (jshu@ti.com) as this DSP software material also requires an NDA.

Regards,

Radio Joe