Hi, what would you suggest for a first Spectrum Analyzer

[MarkoAnte]

Hi,
so my dripping money supply has finally made a puddle big enough for me to invest into a Spectrum Analyzer. What direction would you suggest me going in:

-a old Spectrum Analyzer or
-something like this Tektronix 1401A or some other plug (does somebody still make this) or
-a new rigol DSA815

Would you get an older Spectrum Analyzer for the better quality (I have a OWON scope and a tek TDS 420 and I like using my tek way more) or do you just get more bang for your buck from a rigol DSA815.
1,5 Ghz is more then enough for my needs, tracking gen would be nice also, but not 100% needed.
Sorry if this was already discussed somewhere else.

[KJDS]

What do you want to use it for?

[nctnico]

I'd look for an Advantest (with an LCD display) on Ebay.

[MarkoAnte]

I'm a student and a hobbyist  so basically a wide range of applications (at the present building a FM/AM transmitter for my pirate radio station jarharhar   ) but I'm certain that I would not brake the 1,5 ghz mark of the rigol unit and I would be nice to have the lover bandwidth limit also low - but I can always use my FTT on my tek scope.
Are there other stuff that I should be aware when buying one?

[KJDS]

Important parameters are

Amplitude features:-
Noise, which limits the smallest signal you can see.
Dynamic range which limits how you can see large and small signals at the same time.
Linearity.
input damage level. If you're going to be testing a transmitter, put a big attenuator in the path.

Frequency features:-
Frequency accuracy. Many old analyzers were not synthesized, so the frequency is a bit of a guess. Some more recent analyzers only synthesize the mid frequency and sweep round that, so the span is a bit guessed.
Resolution bandwidth, which limits how close two signals can be and seen as separate on the screen, the smaller the better

Whilst a tracking generator can be useful to get an idea of performance, I would far rather use a scalar or vector analyzer for accurate measurements, however I appreciate this isn't always practical. Some reading on measurement uncertainty so you understand the limits is good.

[MarkoAnte]

Should I get something like this a tek 8558B
http://www.surplus-electronics.co.uk/urchin/assets/images/_mod/20111215_163014.7903.209.800.-1.85.jpg
for around 500 euros or should I spend like 1,200 euros on a HP 8594 E or mabey a new rigol DSA815-TG for 1.200 evro.

[kc9qvl]

I would pass on the 8558b

[PaulAm]

Even today, the specs on the HP 8568b are hard to beat.  It will cover 100 Hz to 1.5 GHz with 10Hz resolution.  Very intuitive user interface.

It comes in two pieces, an RF and a display unit.  You can swap out the RF unit for an 8566 which will go to 22 GHz.

On the down side, it's big, it's heavy, it's around 30 years old, you could run into the problem of a dim CRT and buying a working unit will still likely cost around $1000 US.

[fqahmad66]

DSA815-TG and you will be more than happy with it..

[Terabyte2007]

I have the Rigol DSA815-TG, Love it! Highly recommended.

[georges80]

I just went through the process - spent a couple of weeks checking Ebay and other sites to get an idea of what decent HP gear was selling for. Many apparent bargains that would then bid up to $4000+ within the last minutes.

What became clear was that even in the 2nd/3rd hand market to find anything that covers the ~DC to 1.5GHz band AND provides some decent quality and features was going to cost >>$1k. I consider being able to store the scan as well as being able to carry the piece of equipment to the problem important, so antique CRT based boat anchors were out of the picture.

One advantage of following some Ebay bids (especially the ones that are 5 days or more out) is that it gives you time to search for the manual/specs etc and compare. Time to search the web to see what the street prices appear to average at. Time to see what kind of support still exists - especially from forums etc.

In the end for the $ and for the features (modern ones like USB and ethernet) the rigol 815(tg) kept coming up as the top contender. Lots and lots of reviews and youtube videos and NONE finding any flaws/problems. The published specs appear to be accurate and not marketing fluffed and I never found a review that wasn't positive.

So, last week I ordered an 815-tg and then went on the web to order various cables, adapters, jigs and attenuators. So, factor in another $200 - $300 or so above the SA cost for 'accessories' depending on your planned use for the SA. In my case I'm looking to use it initially for EMI/RFI testing and layout/shielding/cabling investigations. For that the 1.5GHz BW is fine. It also covers any of my Ham requirements if/when I get time to play in that area versus all my 'pay for food' engineering.

cheers,
george.

[G0HZU]

In my opinion there are two 'top' contenders in the 1500MHz range.

One is the old heavyweight HP8568B and the other is the Rigol DSA815. They are both 1500MHz spectrum analysers but they represent the ideal analyser for two distinct types of use.

In terms of bang for buck for serious RF design (even serious RF design today) the HP8568B is simply unbeatable. I have one here and it is a fabulous piece of engineering. But it's big heavy and there's a lot of fan noise. It's a big old beast.

It excels for things like the design of VHF/UHF single loop synthesisers as it has very low phase noise close to carrier and also has pretty good performance when measuring noise accurately even in the presence of big signals.

The dynamic range is reasonably good by today's standards and the overall measurement uncertainty is respectable even today.

But few hobbyists need this level of accuracy and this is where the Rigol DSA815 scores heavily. It sacrifices some front end RF performance and LO performance in favour of low cost, small size, low weight and low power consumption. Obviously you can't seriously use it for things like synthesiser design because the LO phase noise is so high but I would expect that the typical user targeted by Rigol for this analyser won't care about LO phase noise anyway. The linearity is 'adequate' for hobby and semi serious design work.

It also exploits modern technology so you get the benefit of a digital IF and decent RBW filters in the DSP and very good log accuracy over a very wide range (hopefully). It also has a modern display and is highly portable and it looks pretty good too.

Plus you get to buy it NEW and with a warranty for the same price you would pay for a reasonably well maintained HP8568B that could be nearly 30 years old...

In my opinion Rigol have scored a marketing bullseye with the Rigol 815. I'd love to have one here to review. I already have several high end spectrum analysers from the 1980s but it would be very interesting to do a proper technical review on the Rigol. As far as I know, no one has done this yet?

[iDevice]

I would recommend the DSA815 as well.
It's an incredible bang for buck entry level DSA.
As long as you understand that even though it's feature rich (and even more if you want), has a gorgeous display and is very compact, it's still an entry level by some of it's basic characteristics that don't even come close to 30 years old professionals SA's.
Phase noise is quite bad, not catastrophic but below standard to the point that you generally measure the phase noise of the DSA instead of the DUT.
DANL isn't great either. You need to use narrow RBW and preamp to the point that sweep time becomes slower that comfortable to achieve decent noise floor.
And talking about sweep time, I think this is the most annoying weakness of this DSA, the display refresh rate is much slower that the fastest sweep time, so forget about trying to capture any short event or glitch.
Any old SA does better in these areas, but the DSA-815 probably beats them flat in accuracy, automated measurements, ease of use, and many other areas where an analog front-end can't match a digital one.

But, an otherwise quite decent DSA like the DS-815 at 1500$ ? No one would have expected that a few years ago, so I think any beginner with that budget at hand should consider it seriously.
When you are just a hobbyist you don't do very serious measurement anyway, so this baby is perfectly adequate and allows you to play and experiment with all of the latest features you can find in more advanced tools.
You won't be able to do that with very old analog SA's.

One suggestion though, I would consider the TG version because for a few bucks more, it gives you a lot more usability and some insight at network analysis.
And you can't add it later, it has to be built in.

[MarkoAnte]

Ty for the great comments guys, really appreciated. 

[G0HZU]

(DSA815) has a gorgeous display

The display certainly does look very nice and puts some fairly recent professional analyser displays to shame

By contrast, a typical HP8568B CRT display will be showing serious signs of ageing by now with uneven focus and a fuzzy graticule at higher brightness settings. Plus the useable CRT area is quite small.

But behind this lies a display engine that outclasses the modern Rigol. In its day the CRT on the HP8568B was a thing of beauty to look at if you set the focus and brightness correctly and looked at the resolution the display offered. I think the Rigol probably only has something like 600x400 pixels inside the 10x10 grid area and the HP8568B has something like 1000x1000 data points inside the 10x10 grid on the CRT.

But put the two analysers side by side today and the HP8568B CRT will usually look small, soft, dull and very tired compared to the high contrast colour display on the Rigol...

In the image below taken from the KE5FX website you can make out the individual digital datapoints in the display as little dots. This is from a dull/worn out CRT that had been restored using a type of cathode zapping method.

There should be 100x100 datapoints inside each little square on the CRT graticule.

A brand new CRT display would have looked much sharper in 1987

[KJDS]

If you jsut want to look at modulation outputs from transmitters, then I have a pile of E4406A vector signal analyzers that do very well in this area. Better phase noise than most analyzers, 0.1Hz minimum resolution bandwidth but no tracking generator and a maximum of 10MHz span. £500 +VAT + shipping if anyone is interested.

[fqahmad66]

As a hobbiest I built a Synthesized RC transmitter using LMX2306. The results are attached. The magenta one is noise. The yellow one is unmodulated and blue one is PPM modulated. The peaks at 10Khz offset are quite visible due to Phase comp.freq. at 10KHz

[G0HZU]

Yes, you can prove a synthesiser is locking on the design frequency with the 815, but because of the high internal phase noise of the instrument it will mask the true noise performance of the synthesiser.

The yellow trace is showing the noise masking limitation of the 815. The purple trace isn't really proving anything.

I assume you’ve added an external attenuator and added an offset because 22dBm is a big signal to feed into the analyser

[bingo600]

I thought a bit aboout getting the Rigol , but ended up getting  a hp 8594E w. Tracking gen. And narrow bw.,gpib and more.    I really like it a lot , but it was almost 4K$. In EU land. Not Many on hours, and excellent crt.

I also Got a hp 4406 vsa, for "zooming"max on a specificere freq. And to do serious phase noise testing.

Those 4406's are great value for the money.

/Bingo

Edit: for light hobby use i'd consider the Rigol , or a hp 8568b , or maybe an advantest.
But if it's crt model , make sure seller sends picts of the crt (burn in & focus).

[fqahmad66]

Yes, you can prove a synthesiser is locking on the design frequency with the 815, but because of the high internal phase noise of the instrument it will mask the true noise performance of the synthesiser.

The yellow trace is showing the noise masking limitation of the 815. The purple trace isn't really proving anything.

I assume you’ve added an external attenuator and added an offset because 22dBm is a big signal to feed into the analyser

I fed it via a 20db attenuator and also some offset. If i remmber correctly the yellow signal did not change much upto 100Hz RB.. so the tx VCO must have been very noisy.

[G0HZU]

I'm not convinced that your 72MHz VCO can be that noisy...

Even with a modest loaded Q of 5 and -10dBm at the VCO the raw VCO phase noise should be better than -105dBc/Hz at a 10kHz offset. Your plot shows about -87dBc/Hz at 10kHz offset and this is the typical noise level of the Rigol LO.

If the Pd freq of your PLL is 10kHz you can kind of predict the system phase noise with a pocket calculator for a low spec design. eg the overall division ratio is 7200 and a typical loop BW would be 500Hz. So the divider noise at 500Hz could be -155 + 20log(7200) = -78dBc/Hz.  Note: The fudge figure of -155 is based on my experience of designing PLLs at around 70MHz with 10kHz PDF. A lot depends on the divider noise floor spec of your PLL chip though...

So by 5kHz it could fall by 20dB/decade to be -100dBc/Hz and by 10kHz the noise will be somewhere around -105dBc/Hz assuming some contribution from the VCO noise.

This is nearly 20dB lower than the Rigol can hope to display because the Rigol is so noisy. So it simply masks the PLL noise with its own noise. I suppose it is possible to design a 72MHz VCO as noisy as the Rigol LO but I'm not sure how many people have ever managed it before

Of course you can do a much more detailed prediction of the PLL performance using something like ADI SimPLL but the above back of a fag packet sums should be pretty close...

The HP8568B first LO covers about 2GHz to 3.5GHz and has very low close in phase noise at the lower end of this range. eg -112dBc/Hz typical noise at 500Hz offset and this makes it ideal for looking at the close in phase noise of VHF/UHF single loop PLLs. i.e. for looking to see if the loop bw and divider noise is as expected. It's LO noise is low/flat out to beyond 50kHz but the best noise performance is within about 5kHz of the carrier. At offsets between 50kHz at 100kHz the HP8568B LO is quite noisy and is of limited use for synthesiser design at these offsets

[fqahmad66]

way back when i made the first pll tx, i tuned it using Agilent equipment at the university lab. The plot for pll tx output with no modulation is attached. will try to measure same with Rigol and post results.

[G0HZU]

Not sure what model of Agilent SA that is... Is it a E4411?

Either way, it isn't wise to try and measure the noise performance of a 72MHz VCO with those analyser settings and with a signal level of -15dBm going into the analyser. I'm not sure what you are trying to demonstrate. If you really can measure the phase noise of your VCO on the Rigol then it just shows that you have somehow managed to make an incredibly noisy VCO at 72MHz

At the end of the day the Rigol has high phase noise. Normally, you would look at your 72MHz VCO on a 2kHz span and look to see how low the noise level is inside the loop bandwidth. With a 10kHz PDF you should be seeing -80dBc/Hz or maybe better at 300Hz offset on a reasonable design. But I think the Rigol will show about -77dBc/Hz right out to 1kHz where a typical VCO in that PLL should be showing about -86dBc/Hz.

You normally aim for the analyser to have phase noise 10dB lower than the noise you are measuring so this would mean you need an analyser with -96dBc/Hz at 1kHz offset. Your Rigol will maybe manage -77dBc/Hz and this is nearly 20dB noisier.

The 30 year old HP8568B manages -112dBc/Hz at this offset. It was designed to have very low phase noise close to carrier to make it good for designing single loop PLLs.

I also Got a hp 4406 vsa, for "zooming"max on a specificere freq. And to do serious phase noise testing.

The HP8568B phase noise is about 20dB better than the 4406 can manage at a 1kHz offset. The 8568B is in a totally different class in this respect to every analyser mentioned on this thread so far

[TerraHertz]

so my dripping money supply has finally made a puddle big enough for me to invest into a Spectrum Analyzer. What direction would you suggest me going in:

Hopefully not too personal a question, but what price limit do you have in mind?
I ask because at some point my own dripping money supply will accumulate enough to afford a decent Spectrum Analyzer, so I'm also interested in the suggestions people make. Better if we all know the practical price range.
Mine is 'lower the better'. Definitely under $1000 for something new, under $500 for 2nd hand.

Now I'm off to pay about $200 just to ship something 'dead, for spares' from the USA. Bah.

[fqahmad66]

Not sure what model of Agilent SA that is... Is it a E4411?

Either way, it isn't wise to try and measure the noise performance of a 72MHz VCO with those analyser settings and with a signal level of -15dBm going into the analyser. I'm not sure what you are trying to demonstrate. If you really can measure the phase noise of your VCO on the Rigol then it just shows that you have somehow managed to make an incredibly noisy VCO at 72MHz

At the end of the day the Rigol has high phase noise. Normally, you would look at your 72MHz VCO on a 2kHz span and look to see how low the noise level is inside the loop bandwidth. With a 10kHz PDF you should be seeing -80dBc/Hz or maybe better at 300Hz offset on a reasonable design. But I think the Rigol will show about -77dBc/Hz right out to 1kHz where a typical VCO in that PLL should be showing about -86dBc/Hz.

You normally aim for the analyser to have phase noise 10dB lower than the noise you are measuring so this would mean you need an analyser with -96dBc/Hz at 1kHz offset. Your Rigol will maybe manage -77dBc/Hz and this is nearly 20dB noisier.

The 30 year old HP8568B manages -112dBc/Hz at this offset. It was designed to have very low phase noise close to carrier to make it good for designing single loop PLLs.

I also Got a hp 4406 vsa, for "zooming"max on a specificere freq. And to do serious phase noise testing.

The HP8568B phase noise is about 20dB better than the 4406 can manage at a 1kHz offset. The 8568B is in a totally different class in this respect to every analyser mentioned on this thread so far

It never occured to me to measure phase noise..just wanted to make my tx clean so not to shoot down fellow's planes. I managed to get that. just wanted to be -55dB down at 20KHz offset. Now everybody including myself shifted to 2.4GHz so no worries.