EVM measurement techniques - RF Cafe Forums

RF Cafe Forums closed its virtual doors in late 2012 mainly due to other social media platforms dominating public commenting venues. RF Cafe Forums began sometime around August of 2003 and was quite well-attended for many years. By 2012, Facebook and Twitter were overwhelmingly dominating online personal interaction, and RF Cafe Forums activity dropped off precipitously. Regardless, there are still lots of great posts in the archive that ware worth looking at. Below are the old forum threads, including responses to the original posts. Here is the full original RF Cafe Forums on Archive.org

-- Amateur Radio

-- Anecdotes, Gripes, & Humor

-- Antennas

-- CAE, CAD, & Software

-- Circuits & Components

-- Employment & Interviews

-- Miscellany

-- Swap Shop

-- Systems

-- Test & Measurement

-- Webmaster

yz

Post subject: EVM measurements Posted: Wed Aug 10, 2005 1:25 am

hello friends,

Is there any techniques to measure EVM (error vector magnitude) with network analyzer or spectrum ?

(Instead of the known technique with vector signal analyzer).

Thanks,

YZ

Top

Guest

Post subject: Posted: Wed Aug 10, 2005 9:11 am

NO

Top

guest

Post subject: Posted: Wed Aug 31, 2005 4:54 am

Hi.I think you can use VNA to measure the CW AM/AM and AM/PM. There is some relationship for them with the EVM. And in ADS you can simulate it with the measured CW AM/AM and AM/PM to get the simulared EVM value.

Top

IR

Post subject: Posted: Thu Sep 01, 2005 2:49 pm

Site Admin

Joined: Mon Jun 27, 2005 2:02 pm

Posts: 373

Location: Germany

Hello,

[/b]This is partially true[b]

EVM is a function of more than one parameter: Besides linearity, it is also depends on phase noise. You can't get information about phase noise from AM-AM and AM-PM measurements.

These measurements solely meant to characterize the compression characteristics of an amplifier/transmitter/system and to get information about the linearity. You still need to get the phase noise information (By doing that measurements with Spectrum Analyzer).

Then when you have these quantities there are models to calculate the EVM. As far as I know, the only method to measure EVM is with Vector Signal Analyzer.

_________________

Best regards,

- IR

Top

Guest

Post subject: Posted: Sun Sep 04, 2005 11:35 pm

Hi...

I understand what you said because I often encounter that I have good AM/AM and AM/PM charateristic but poor EVM ( I am designing the WLANPA now).

However, I found that usually when I added some different large bypass capacitor, the EVM can be improved with even the same AM/AM and AM/PM chracteristic. I don't know if it is what you said the phase noise being reduced. I guess that could be some kinds of low frequency noise. Could you please teach me more about the phase noise and how to reduce it? The only way I know is to add different kinds of large bypass capacitors.

Top

IR

Post subject: Posted: Sun Sep 04, 2005 11:55 pm

Site Admin

Joined: Mon Jun 27, 2005 2:02 pm

Posts: 373

Location: Germany

Hello,

Different types of bypass capacitors can improve many things in a PA circuit, like improve the stability. Concerning improvement of EVM, I don't know if that can do much. What can do more are the matching networks. You should stick to the recommendations of the transistor's/device manufacturer by using the same values of external components in your circuit. On top of that circuit, you can make additional improvements. Use ADS or a similar tool to analyze your circuit (Provided that you have the right model for the transistor), and see what you can do to improve EVM.

Good luck!

_________________

Best regards,

- IR

Top

IR

Post subject: Posted: Mon Sep 05, 2005 8:57 am

Site Admin

Joined: Mon Jun 27, 2005 2:02 pm

Posts: 373

Location: Germany

Hello,

One more thing I wanted to add: Bypass capacitors and bypass networks of the PA (Which often consist from resistos, beads and capacitors) can have an effect on a phenomenon called "Memory Effect", which is a change in the AM-AM and AM-PM curves over temperature and frequency. The "Memory Effect" is a distortion by itself and also affects the EVM.

Why won't you use some sort of linearizer? This will improve your distortion by all means (EVM and other non-linearities)?

Good luck

_________________

Best regards,

- IR

Top

Guest

Post subject: Posted: Mon Sep 05, 2005 10:21 pm

Hi~~

Thank you for so good suggestion. I am a WLANPA IC designer so I cannot use linearizer or specified bypass network because I have to shrink my chip size as small as possible and also avoid external matching components as few as possible. Also I cannot follow the suggested external matching values on the datasheet because I will write the datasheet ^_^

Actually, I think the key point of good linearity is the adjustment of on-chip active bias circuit beside the RF matching of the whole circuit. However, both of the bias circuit adjustment and the RF matching are the causes of AM/AM and AM/PM behavior what we will get. But the phase noise as you said might not be related to them but should be related to the bypass capacitors. In fact I ever encounter one interesting thing. When I changed my 0.1uF bypass capacitor size at Vcc from 0201 size to 0603 size, the EVM improve from 4% to 3% and all other RF performance kept the same! Also I am interested in the memory effect you said. I guess that can be concluded to the time variant AM/AM and AM/PM behavior and will determine the temperature deviation of the RF performances.

Top

IR

Post subject: Posted: Mon Sep 05, 2005 11:44 pm

Site Admin

Joined: Mon Jun 27, 2005 2:02 pm

Posts: 373

Location: Germany

Hello,

All those things are related to one another. The size of capacitor changes its Q due to different resistance. You also change its resonance frequency and its parasitic inductance. unsufficient bypassing can cause AM or FM modulation over the signal - phase noise.

Active bias is a good way to control over your AM-AM and AM-PM curvers. As you may know, once you provide a low bias to the PA, you will get a curve with as bigger dip (at the average region of the signal) and the amplifier will enter more quickly into saturation. The opposite will happen with a higher bias. Higher bias provides therefore a more linear and less power-efficient amplifier and lower bias provide the opposite (per same output power).

_________________

Best regards,

- IR

Top

Guest

Post subject: Posted: Tue Sep 06, 2005 12:49 am

Hi~~

Yes this kind of phenomena at low bias we call it "EVM plateau" and usually it is caused by high AM/AM behavior. However, recently it's found even we have high AM/AM behavior for CW operation, the momentary AM/AM behavior for OFDM signal operation could be very flat and the EVM plateau can keep very low. So, we can get very good PAE but also very good EVM. I guess you may know it and it is called " IMD cancellation". However, until now it is hard to simulate and design it because if you take so high CW AM/AM into ADS simulation you will get poor EVM performance. Actually I doubt that the CW AM/PM behavior

might help us to see if IMD cancellation happen but I've not yet verified it.

Maybe in the future this kind of phenomena can be realized more and we can easily design this kind of very linear and very high efficient PA.

Top

SKChen

Post subject: Posted: Tue Sep 06, 2005 1:14 am

Lieutenant

Joined: Tue Sep 06, 2005 12:58 am

Posts: 3

Location: Taiwan

Hi...

I am the Guest as above and I decided to register an account. I am now working for Japanese to design the WLAN PAIC. When I was in my graduate school, I study the feedforward linearization method of PA. But when I joined my company I found that many linearization method are not available because we have very limited cost issue and size issue. So we have to think about more for the on-chip active bias circuit. I've ever heard that in Israel there is some company which can integrate the feedforward linearization into a chip. I don't know if it is real because I think it should be very difficult to make the coupler on chip.

Top

IR

Post subject: Posted: Tue Sep 06, 2005 2:26 am

Site Admin

Joined: Mon Jun 27, 2005 2:02 pm

Posts: 373

Location: Germany

Hi SKCHen,

First it is good that you made an account!

Feedforward will always require an additional PA (The Error PA) and also couplers and cancellation networks and I don't think that it is achievable with today's technologies.

Nowadays, Feedforward and more advanced methods are available at the module level and not at the chip level.

There are however today ASICs that are implementing Digital Pre-Distortion algorithms for RF PA's and also chips that are doing Analog Pre-Distortion.

_________________

Best regards,

- IR

Top

SKChen

Post subject: Posted: Tue Sep 06, 2005 5:09 am

Lieutenant

Joined: Tue Sep 06, 2005 12:58 am

Posts: 3

Location: Taiwan

Hi IR:

Yes in the past years the Digital Predistortion PA research rapidly grows up but unfortunately for us RF engineers it's difficult to enter that field especially when we don't have the baseband talent in our company. So, we have to try to find more tricks in our circuit itself. But I guess that in the future we RF engineer may need to cooperate more and more with the baseband engineer since I've heard that TI has already combined the baseband part and RFIC part of GSM mobile phone into a single chip by DSP. The only things remained outside for the traditional RF engineers may be only PA, Switch, and antenna. But anyway, some kinds of physical level things will not disappear. People still cannot use digital circuit to realize the PA, switch, or antenna. Nevertheless, we RF engineers have to estabilish our knowledge background more like this topic, how is the EVM relationship with the traditional RF performance items like AM/AM, AM/PM, phase noise as we know so far. That's why I visit websites like here to find the answers because there is almost no book that can tell us about topics like this.

Top

IR

Post subject: Posted: Tue Sep 06, 2005 5:47 am

Site Admin

Joined: Mon Jun 27, 2005 2:02 pm

Posts: 373

Location: Germany

Hello SKChen,

Yes that is true, the cooperation between the RF Engineer and Baseband Engineer has to grow due to the interaction of the linearization issue.

I have worked with several linearization methods but all were intended for PA modules and not chips. Nevertheless, I saw and seeing now (In my current job) the deep meaning and significance of high-speed A/D and D/A converters, issues of phase noise, jitter, dynamic range of A/D and D/A and their impact over the system performance.

If you would provide your email address I will be happy to write you more.

Good luck.

_________________

Best regards,

- IR

Top

SKChen

Post subject: Posted: Tue Sep 06, 2005 9:12 pm

Lieutenant

Joined: Tue Sep 06, 2005 12:58 am

Posts: 3

Location: Taiwan

Hi IR:

Yes I also heard about that the high speed ADC or DAC help DSP people can handle many things at baseband level which originally should be handled at RF level. Embarrassingly I am very unfamiliar to that since in my office we only desing PA and switch. However I still want to learn more. I've made my mail address visible so if you have leisure please pass me some information. I do appreciate your kindness.

Posted  11/12/2012