Hi all,

I am trying to 'reverse engineer' block level performance requirements from the EVM and ACLR specifications for a high power OFDM transmitter, for LTE.

I'm aware of the rough relation between EVM and its various causes (non-linearity, phase noise, IQ phase and gain error and carrier leakage). But, the trouble I'm facing is in translating and partitioning (downwards) the EVM specs. into block (Mixer, PA, Filter etc.) level requirements for an LTE transmitter.

Some of the questions I have are:
-how much of each of the blocks do I need to know before I can reverse engineer their performance numbers, from the specs.?
-how much (percentage) of a particular performance to give each block? As for the ACLR specs., I'm assuming that the biggest culprit is the non-linear PA. Again, the issue here is to distribute the ACLR (i.e. nonlinearity) budget among the various transmitter blocks.

If you have any advise or comments, that I could use, I'd be happy to hear form you!

Thanks in advance.

Cheers!

@ RFICDUDE

Thanks for the comments! I probably understand roughly, the individual mechanisms and how they might contribute to the EVM and ACLR degradations, and that is, to say, only a part of the solution.

To follow up, and determine block level performance reqmts., I need to distribute those contributions among their 'contributors' in the Tx chain. i.e. I know that the EVM limit for OFDMA with QPSK is 17.5% (for LTE). There's a following step which determines how much comes from carrier leakage, from phase noise , non-linearity, IQ mismatch etc.. But, how do I determine how much of that to attribute to the amplifier, mixer, etc (contributing to IQ mismatch, non linearity effects on the EVM).

A long question, but I hope you get the drift...!!

The problem is I'm not sure how to do that (from not having the experience!!).

Any renowned/interesting papers that you can share, would help.

Also, I'm trying to trace this atricle - Michael Leffel, "Intermodulation Distortion in a Multi-signal Environment," RF Design Magazine, June 1995, pp. 78-84. ... if you've come across is before?!

Oh, I forgot to mention an overwhelming complication these days when it comes to OFDM systems. For WLAN/WiMAX the EVM of the received signal depends strongly on the sophistication of the equalization filter used in the receiver. The unfortunate outcome of this is that EVM degradation is worse than one would expect for some static errors such as image rejection (gain/phase imbalance). An even more tragic result is that the specification on transmitter EVM is usually tightened up to make up the difference. I have seen system designers design for 3dB tighter image rejection to account for this.
I hope LTE is a little more definitive about what the receiver equalizer assumptions are for system measurements on the transmitter.
------
Distributing the budget really depends on your knowledge of the blocks and the architecture. A rough distribution can be guessed at by simply assigning contributions logically to blocks that you know are going to impact certain performance parameters which impact EVM such as gain and phase imbalance, distortion, phase noise, LO leakage (depends on if LO falls within the Rx bandwidth), etc.

Gain imbalance is most often atributed to the baseband amplifier/filter circuits all the way up to the mixer. Slight LO amplitude imbalances may not be significant if the mixer limits the LO signal (Gilbert 4-quadrant mixer).

Phase imbalance is relates most directly to the quadrature LO balance.

You can budget gain and phase imbalance by using image rejection as the measure of how these two contributions degrade EVM.

LO leakage relates directly to DC offsets in the baseband right up to the mixer. However, at really low output power the LO leakage through capacitive and/or inductive coupling paths will dominate the low level leakage.

Distortion is usually budgeted to allow the PA to dominate the spectral mask, so the transmitter before the PA usually has a roughly 10dB tougher spec than the PA for adjacent channel distortion. However, with a tight EVM budget the transmitter distortion may have to be even less to minimize any degradation to EVM at high output power.

You also have to be willing to analyze the EVM over the power control range because the distortion contribution will generally reduce as power is reduced (depends on how gain and the transmitter and PA output sections are controlled). This permits other contributors to take up some of the distortion budge when distortion is not a significant contributor.

Roughly, the inband distortion (dBc) will be worse than the out of band distortion by a couple of dB for high peak to average signals like OFDM. Simple voice only WCDMA signals may have the same or less inband distortion than out of band distortion. So, unfortunately, the impact of distortion depends on the waveform mode the transmitter is operating in.

The specifics of all of these (I’m sure this is what you are looking for), are unfortunately dependent on the architecture being used and the knowledge of block level performance. Therefore, filling out the initial sheet is a matter of guessing and honing the numbers against a design team that is willing to say what is realistic or not (i.e. what are they willing to bet their paychecks on).