Laser bandwidth

[jazzalex]

Hello all,

I am currently in a fundamental discussion with somebody regarding the carrier bandwidth of a microwave generator vs. lasers. In that context there are two open questions:

1.) Is it correct that with currently available high-quality generators (such as Anritsu, R&S etc.) one can assume that a microwave carrier of e.g. 9 GHz does not exceed a bandwidth of approx. 1 kHz (or even less) ?

2.) Is it right that lasers have a significantly higher bandwidth of (much) more than 1 MHz ? If yes, what is the currently available lowest bandwidth of high-quality (high-cost) devices ?

Thanks in advance,
best

Alex

[wkb]

Not too sure if I get your question correctly but:  100GbE Ethernet is driven with semiconductor lasers.

[jazzalex]

Ahhhh - sorry, I am not talking about the throughput of a laser driven device: I am talking about the spectral bandwidth of the laser. Any hint appreciated.

Thanks

[IanB]

I'm not a communication or information systems engineer, but I think your questions are not posed in a way consistent with theory. This would make it difficult to give the kind of answer you are seeking.

Firstly bandwidth in an information sense is usually expressed in terms of information transfer, in bits per second.

When a data stream is transmitted over an analog link, such as laser, or microwave, or even just copper wire, this data stream has to be modulated onto a carrier frequency. Data is transmitted as symbols per second, where each symbol can carry one or more bits. Higher frequency carriers can support more bits per symbol and/or more symbols per second, allowing more bits per second, and higher bandwidth.

The symbols have to be encoded onto the carrier wave by modulation. If you had a pure carrier wave of a single frequency with no distortion, then it would be carrying no information and would therefore have a bandwidth of zero. Naturally, as a pure wave it would also have no spread of frequencies, making the spectral bandwidth zero Hz as well.

As soon as you modulate the wave to carry information, you make it deviate from a pure single frequency--you introduce distortion, harmonics, other frequencies mixed in, however you want to describe it. So simultaneously with making the wave carry digital information (information bandwidth), you also make the frequency spread out from the pure carrier frequency (spectral bandwidth). The two are connected. And the observed bandwidth depends on what you do to the carrier to modulate it.

A related issue how much information can you actually squeeze onto the carrier? What is the maximum limit, given some practical constraints? In such cases it is nearly always true that a higher frequency carrier can be made to carry more information than a lower frequency carrier. So lasers can carry more information then microwave links. Lasers have higher bandwidth.

[jazzalex]

Thanks for that reply ! I understand your point and what you write is of course 100% correct. However -- even without modulating any information -- a laser itself already has a spectral bandwidth, and the amount of bandwidth I know e.g. for the special case of a Helium-Neon-Laser (as attached) but not of a fiber-laser in conventional telecommunication. If anyone does, please let me know.

Best

Alex

[Jad.z]

Optical systems use carrier frequencies in the range of (2 X 10^14) Hz

Communication Systems - A. Bruce Carlson, Paul B. Crilly, Janet C. Rutledge.4th ed.    Page: 102~106

[IanB]

Thanks for that reply ! I understand your point and what you write is of course 100% correct. However -- even without modulating any information -- a laser itself already has a spectral bandwidth, and the amount of bandwidth I know e.g. for the special case of a Helium-Neon-Laser (as attached) but not of a fiber-laser in conventional telecommunication. If anyone does, please let me know.

This article may help: http://www.rp-photonics.com/linewidth.html

The important term apparently is not bandwidth but linewidth.

[jazzalex]

This article may help: http://www.rp-photonics.com/linewidth.html
The important term apparently is not bandwidth but linewidth.

Right on ! That indeed helps and yes: Now I know the precise term to use !

Thanks a lot

Alex

[ejeffrey]

Linewidth is an attempt to give a single number to the phase deviation of a laser, and as such you have to be rather careful with its applicability.

Many lasers are not single mode.  This means that their frequency spectrum looks like a comb, although usually with only a few teeth.  Obviously then you have to be careful about what you call the linewidth.  Diode lasers on the other hand normally operate on a single mode, but unless they are treated gently they have a habit of hopping back and forth between two adjacent modes.  For slow measurements this looks somewhat like multi-mode behavior only more objectionable.

Furthermore, linewidth (defined as the full-width half-max in the spectrum) is not the full story even for a single mode laser.   Very often we are concerned about the spectral content way, way outside the linewidth.  For instance, if you have a laser with a 100 kHz linewidth, but you are modulating it with a 100 MHz signal, you actually care about the noise content from the laser 100 MHz away from the carrier.  For sure this is low, but is it 80 dB below the carrier or 130 dB?  That affects your final SNR.

OK, with all the disclaimers out of the way.  The best free-running lasers are monolithic non-planar ring lasers (a type of Nd+:YAG laser) and some fiber ring lasers.  Good ones of both have an intrinsic linewidth of around 1 kHz.  General purpose solid state lasers have a typical linewidth in the MHz range.  Diode lasers run a huge range, but are generally quite broad. 10s of MHz or 100s of MHz are not uncommon.  With house training (using an external grating for optical feedback) a diode laser can have its linewidth reduced to 100 kHz or so.

Any of the above lasers can be frequency locked to an external cavity or other reference.  This can reduce their linewidth to the sub Hz range, or even lower.  However, this only suppresses the noise within a certain band around the carrier.  If you take a 100 kHz diode laser and lock it to a reference cavity you can reduce the linewidth to near zero but you only suppress the phase noise out to around 1 or 10 MHz.  This laser will still have plenty of noise at a 1 GHz offset from the carrier.

[jazzalex]

Allright -- this info is even more useful. The background is the following: I have a research project going, in which I AM-modulating signals of max. 1 MHz onto the carrier and the carrier itself has to be as narrow as possible in terms of bandwidth. This project is currently applied in the microwave domain and for certain reasons I want to move to the fiberoptic domain. As a next step I need to figure what equipment I need (and how expensive it is).   

Any of the above lasers can be frequency locked to an external cavity or other reference.  This can reduce their linewidth to the sub Hz range, or even lower.  However, this only suppresses the noise within a certain band around the carrier.  If you take a 100 kHz diode laser and lock it to a reference cavity you can reduce the linewidth to near zero but you only suppress the phase noise out to around 1 or 10 MHz.  This laser will still have plenty of noise at a 1 GHz offset from the carrier.

That noise I wouldn't care about but in that context I have another question: With that being said I am wondering if that would also work for very cheap diode lasers as we find them in standard telecommunication gear ?

Thanks

Alex


 
   

[ejeffrey]

Linewidth is usually a measurement of phase noise, excluding amplitude noise.  If you are doing AM modulation it is likely that you care about amplitude noise more than phase noise, although the phase noise will still contribute to the occupied bandwidth and the susceptibility to dispersion.  Once you get your linewidth smaller than the modulation bandwidth you should be good to go.  In amplitude, most solid state lasers are intrinsically very low noise except around their relaxation oscillation frequency.  Low frequency noise is due instead to external sources such as mechanical vibrations, pump fluctuations, and unwanted optical feedback from backreflections.  In particular, the worst cause of amplitude noise in diode lasers is the injection current.  If you have a low noise current source you have a low noise diode laser.

Anyway, I am not an expert on telecom band lasers, but here is where I would start.  For turn-key systems you can look at the Thorlabs INTUN lasers, the Newport/New Focus Velocity lasers, and the Toptica DL systems.  They will give you a nice fiber coupled laser that is reasonably narrow (in the 100-300 kHz range) ready to be locked to an external reference cavity.  Be prepared to pay a lot for the convenience.  For a bit more flexibility, you can look at the Thorlabs tunable laser kits.  They are also a bit cheaper.    These systems are all relatively expensive because they are tunable, I expect that you can get narrow linewidth fixed wavelength lasers cheaper, I just don't know where.  I did find a single frequency diode laser chip from Thorlabs:

http://www.thorlabs.de/NewGroupPage9.cfm?ObjectGroup_ID=4907

That is claimed to have a linewidth of < 100 kHz.  Of course, you need to put a good temperature controlled mount around that and get a low-noise current driver.  Again, you would probably have to lock to an external reference to get down below 100 kHz.

I don't think you will be able to use the super-cheap "fabry perot" laser diodes and get their linewidth down to 100 kHz easily.  Their intrinsic linewidth is just too great to effectively control through electronic feedback.  The normal way to deal with this is to AR coat one facet of the laser chip and use an external grating to generate a bit of optical feedback.  If done properly this reduces the bandwidth to below 1 MHz and makes the laser tunable.  This is how the commercial tunable diode lasers work.

The enemy of all narrow linewidth lasers is uncontrolled optical feedback.  Every optical element in the beam will reflect a fraction of a percent or more back towards the laser.  This optical feedback drives them crazy since it is very sensitive to vibrations and temperature drifts.  This is why you need at least one if not two stages of optical isolators before you start putting in things like reference cavities.

[jazzalex]

Awesome ! Nothing else left to ask -- at least for now. I will make a start and get back once new questions/ideas come up.

Thanks for your help -- this is a great forum !

Alex

[Wim_L]

Thanks for that reply ! I understand your point and what you write is of course 100% correct. However -- even without modulating any information -- a laser itself already has a spectral bandwidth, and the amount of bandwidth I know e.g. for the special case of a Helium-Neon-Laser (as attached) but not of a fiber-laser in conventional telecommunication. If anyone does, please let me know.

Best

Alex

Knowing something about HeNe lasers, I do believe that spectrum you posted tells you a lot more about the resolving power of the spectrometer you used than it does about the HeNe laser linewidth (which is far too small to measure with normal general-purpose spectrometers). Typical general-purpose spectrometers are limited by monochromators with about 1nm bandwidth, and measuring laser linewidth with those is only useful for fast pulsed lasers (say, about 10nm FWHM on the spectrum for a 100fs pulse).

[johnmx]

Usually the lasers for telecommunications have a linewidth higher than 100kHz. 1kHz is for medical purposes.

[w2aew]

Even the lasers used for today's coherent optical modulation schemes are typically 100kHz or more inherent linewidth.  Of course with coherent modulation bandwidths in use that are 15-20GHz, who cares!