Frequency counter. Do you really need one?

[ben_r_]

I guess Im a little confused on the matter, but whats the difference between a dedicated frequency counter and the frequency measurement setting in a DMM? Is it just that a dedicated unit usually has a higher and/or lower range? Is a dedicated more accurate?

[PA4TIM]

The frequency measurement capabilitys of a DMM are very limitted.
Most are not very sensitive, no triggering options and low resolution.
Most couters are 8 or more digits and even 40 year old models go to 50 or 60 MHz.  ( my old HP and Philips both with nixies go to 60 MHz, my old LED Tek is 100 MHz, my Philips is 1 GHz, i have a 17 GHz counter and some homebuild ( 50 MHz and 2,5GHz)

It is very easy to build a counter. You need a uP, 2X16 digit LCD, one or two logic ICs and a prescaler and you can measure over 1 GHz.

But a commercial one can do much more. External reference connection, period measurement, event counting , frequency, difference between two signals, attenuators, filters. Some can even measure things like rpm.
triggerlevel and filters is often needed is signals are not very clean.

[ben_r_]

The frequency measurement capabilitys of a DMM are very limitted.
Most are not very sensitive, no triggering options and low resolution.
Most couters are 8 or more digits and even 40 year old models go to 50 or 60 MHz.  ( my old HP and Philips both with nixies go to 60 MHz, my old LED Tek is 100 MHz, my Philips is 1 GHz, i have a 17 GHz counter and some homebuild ( 50 MHz and 2,5GHz)

It is very easy to build a counter. You need a uP, 2X16 digit LCD, one or two logic ICs and a prescaler and you can measure over 1 GHz.

But a commercial one can do much more. External reference connection, period measurement, event counting , frequency, difference between two signals, attenuators, filters. Some can even measure things like rpm.
triggerlevel and filters is often needed is signals are not very clean.

I see. Makes sense. Ive had my eye on picking up a used Agilent 53181A for a bit now. Perhaps Ill grab one the next time I see a good deal. Be nice just to have one around. Thanks.

[saturation]

A quality frequency counter provides high accuracy, in ppb.  Whatever feature it has for measurements e.g. period, frequency, rise time, fall time, etc., can also be made with lower cost counters, or even your DSO, but not with the same accuracy.  A  good counter will have at least 10 digits of resolution or greater, to give you 1 ppb accuracy, thus it can resolve a 10 MHz signal down to 0.01Hz.

< $200 ones go to about 2-3 Ghz easily,  accuracy in ppm, stable to 8 digits.  You can get many of these handheld.



Why you need ppb accuracy is another story!  but ppb separates costly from cheaper counters.

[G0HZU]

I would say a modern quality counter will also have a much faster refresh rate than those TTi counters. The TTI 930 bench meter looks like it needs to wait 100 seconds (!) to display to 10 digits. That is very slow. These days a decent reciprocal counter should be at least 100 times faster than that when displaying 10 digits.

Why you need ppb accuracy is another story! 

In my case I need to be able to measure/set the 10MHz ovens in my test gear to a fairly decent accuracy. Eg the spec when aligning the 10MHz reference on one of my analysers is to set it to +/-0.001Hz at 10MHz after a 48hour warm up. Obviously I don't try for that degree of accuracy but I do like to keep the various ovens within 0.1Hz of 10MHz. I try to keep the error below 0.05Hz and some of them manage <0.02Hz error over several months.

This may seem OTT to many, but some of the test gear runs up at many GHz so the errors soon mount up especially when mixing two microwave signals down to baseband. eg a 1Hz error at 10MHz means a 1kHz error at 10GHz.

[dmmartindale]

I would say a modern quality counter will also have a much faster refresh rate than those TTi counters. The TTI 930 bench meter looks like it needs to wait 100 seconds (!) to display to 10 digits. That is very slow. These days a decent reciprocal counter should be at least 100 times faster than that when displaying 10 digits.

Actually, the TTi provides pretty good resolution for a reciprocal counter.  Reciprocal counters measure for a number of complete input signal cycles, but that period includes one incomplete timebase cycle at the start and another at the end of the gate time.  If the timebase frequency is 10 MHz and the gate time is 1 s, then the measured period is always going to be about 10 million cycles plus or minus 1 cycle - and that's good for only 7 digits of resolution.  (B&K 1823A or 1856D are examples).  The TTI multiplies the 10 MHz timebase up to 50 MHz for counting, and so it manages to do 8 digits of resolution (+- 2 counts in the LSB) in 1 second.  But there is still 20 ns uncertainty in the length of the measuring period, and thus 2 parts in 10^8 uncertainty in the frequency.

You're probably thinking of counters which also have analog interpolator circuits that measure the length of the first and last incomplete cycles of the timebase when the gate opens and closes.  This can provide much better precision for the start and end time, and thus several extra digits of output in 1 second gate time.  If you see a counter providing 10 or 12 digits/second, it is likely an interpolating counter - not a reciprocal counter with a really high timebase frequency.  But interpolating counters are still rather expensive new.

And the TTi does have one rather interesting feature: It is a "capture and continue" counter, where the counting registers are saved but not reset each time they are sampled.  With no "dead time" or lost cycles, the counter can provide an update rate faster than the gate time.  With a 10 s gate time, the display updates every second, but each value is based on measurements 10 seconds apart using a "rolling window" on saved measurments.  With a 100 s gate time, the display updates every 2 seconds, while each value is based on the most recent 100 seconds of measurement.

- Dave

[babysitter]

Shameless reusing of some advice I gave before regarding counter selection, edited a bit to solve the dependency of what was going on in that thread:

Do you want to count frequencies above what you can use a scope for, or do you want way higher resolution that possible with peak-counting on the scope?
Do you want to measure one or several events that are too slow for convenient o-scope-peak-counting?
Do you want to log frequencies or time-intervals?

If you answer at least one of those with yes, then go for Universal counter.
Consider using a wide-band radio receiver as an alternative for high frequency stuff, you might be able to listen into your signal. Google for SDR dongle.

If you want to do digital stuff, consider a Logic analyzer (OpenBench LA for example), you can read the frequency and time intervals with that device, too.

If you want the convenience of connect to DUT and read frequency, go for universal counter.

If you only want frequency reading, a simple frequency counter is enough, a universal counter will help if you want to compare frequencies or have exotic trigger/arming requirements for example.

About everything you can do with a UC you can do with other stuff, e.g. abusing the LO of a superhet radio as signal source, by setting the radio at one IF above the frequency where you want a carrier on air to test a receiver. Wideband radios, Lissajous method on cheaposcope are available too. But without the convenient number readout.

I like my Box 5334A, but there are other from HP and there are also other manufacturers of course who make decent rigs.

What upper frequency limit do you look for ?

Check your space constraints, means check what fits best on your desk.

Get the manuals matching the offers on ebay for example from KO4BBs website, compare specifications.
A bad reference oscillator is not a no-go - you can replace it with something better, which is exactly what I did.
A defect box is always a gamble, mine had defects not listed but was way cheaper than the others.
"only" had series resistors in input stages burnt, easy to replace.

[TunerSandwich]

One could use a frequency counter to verify that their scope count was accurate. If you can afford it I don't see how it could hurt. If you work on designs that use timing critical components....it can never hurt to get a second opinion.

[dom0]

IF one wants to buy / build a frequency counter here's what I would watch for:

- No counters with stupid 10 MHz references. They're slooooow. Remember that in a "modern" reciprocal counter measurement speed is the log10 of the greater frequency (ref / input). E.g. a stupid reciprocal counter with a 10 MHz reference measures a 100 kHz only with 7 digits per second, i.e. 6 digits if you want 10 updates per second.
- Many cheap counters are build like this. Most DIY counters do it like this. Mine uses a 20 MHz ref which is only marginally better. Next one (if I ever come around to design it) will probably use some FPGA or CPLD and a synthesized/PLL'd poly-phase reference ; much faster.
- Better counters still of course use a standard 10 MHz reference, but generate a much higher internal reference clock from it. See next item in the list...
- Good older counters are sometimes still good. For example the HP 5345A (from the early seventies, as seen in Days Of Future Past) uses a 500 MHz reference (check out their schematics - they use resonance amplifiers locked on harmonics of the 10 MHz reference clock, absolutely amazing circuit technique!) and thus achieves something like 8.7 digits/second.
- Analog front end matters. For many uses a counter which can't measure low level (<100 mVrms) signals is simply useless.
- You want a high-Z input good for at least 50 MHz
- You want adjustable trigger levels
- You probably want a front end which can trigger on narrow pulses
- Consider counters with interpolation, they don't have the usual +-1 gate uncertainty and interpolation usually gives 1 digit/second more meas speed
- For some uses a TIA is to much more useful. TIAs are, however, much more expensive and rare and you really wouldn't ask if you didn't need it (but if you do, consider Yokogawa TIAs, they're much cheaper than the HP ones and have excellent performance, just like the HP ones — also, for a DIY TIA you really want a top-notch analog front end with the least amount of trigger uncertainty (i.e. jitter) you can get)
- Oh, there's also Spectracom/Pendulum. But chances are that if you find someone selling one of their counters the seller knows what they are really worth ("Worth their weight in gold" - literally). Pendulum split from Philips/Fluke and all their stuff is top-notch.

[G0HZU]

You're probably thinking of counters which also have analog interpolator circuits that measure the length of the first and last incomplete cycles of the timebase when the gate opens and closes.  This can provide much better precision for the start and end time, and thus several extra digits of output in 1 second gate time.  If you see a counter providing 10 or 12 digits/second, it is likely an interpolating counter - not a reciprocal counter with a really high timebase frequency.

Yes, I'm thinking along the lines of my old Anritsu counter. It's still described by Anritsu as a reciprocal counter but they use reciprocal /vernier techniques to enhance the resolution. It works really well for my needs at 10 digits per second although modern counters outclass it. I assume it uses digital interpolation rather than analogue but I've never really studied its operation due to the lack of a manual.

..TBH I'm not really that 'into' measuring frequency very accurately. I only bought the Anritsu because it was so cheap on ebay. I saw it with a few minutes left on the auction and put in a low bid and won it for just over £60.

Otherwise, I think I would have (happily) just carried on using a scope and my homemade offair standard to align my various 10MHz OCXOs to be in reasonable agreement

[Lightages]

if you want a frequency counter, cheap, and accurate, then there seems to be a good option. I stuck my neck out and purchased a Victor VC3165. Before I did I checked for reviews and opinions on it. Places like eham.net and some other places had good opinions on it. For around $100 it is a good deal. If you buy one and it arrives in one piece then you get an oven stabilized clock referenced counter that seems to cover most people's needs. The plastics used in the case are really bad and some people have reported receiving something that sounds like everything is trashed inside because the transformer broke loose. Mine arrived in fairly good shape and it seems to be accurate. I would buy one again.

[Lightages]

The decision is easy. If you need to measure frequencies over 120MHz, forget the PM6671. If you want a better instrument with more options but don't need over 120MHz, then the PM6671.

In reality, most people have some frequency measurement capability with their multimeters. If you have an oscilloscope then you have even a higher range. The VC3165 is probably the best bet for most people IMHO.

[Smith]

I once tried to calibrate an old function generator with a 1GHz scope without HW counter. Could only take a bit acurate frequency measurements at high sample memory, so friggin slow. Drove me crazy. Took out an old HP counter, and was done in no time.

Unless you have a brand new scope with HW counter a seperate counter is much cheaper than replace your scope. Probably more accurate too.

[VK5RC]

Some of the good 'interval timer then invert'  type frequency counters have no down time between timing cycles improving accuracy considerably, esp if you are getting all time-nut - ish and doing Allan deviation etc.
Once you compare the result between a reasonable dso and a good counter locked to a trusted time base,  you become pretty sceptical about some dso and also dmm counters.

[nowlan]

This guy does a review of some chinese frequency counter thing. Seems to perform very well. Not sure when I'd need out outside of RF.

[dom0]

Some of the good 'interval timer then invert'  type frequency counters have no down time between timing cycles improving accuracy considerably, esp if you are getting all time-nut - ish and doing Allan deviation etc.

How is that? They should have at least one cycle of either reference or signal dead time. The only technology I am aware of that can do measurements with no dead time at all is continuous time stamping.

[VK5RC]

If I understand it correctly the HP 5371 was one of the earliest to do this, it doesn't count cycles over a defined time, it times a defined number of cycles, then does the maths to calculate frequency, but can do this while the next timing interval is progressing. See HP Journal below.
http://www.keysight.com/en/pd-1000001409%3Aepsg%3Apro-pn-5371A/frequency-and-time-interval-analyzer?cc=US&lc=eng
http://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/1989-02.pdf
Again if I understand correctly, most good ~ HF range frequency counters e.g. Racal, HP etc (really inverted interval measurement)  mostly use this technique since then (~1989).

[dom0]

That technique is continuous time stamping.


http://www.spectracomcorp.com/Desktopmodules/Bring2Mind/DMX/Download.aspx?EntryId=446&PortalId=0

[VK5RC]

Ah Not heard/noted that term before, thanks for the reference,  the HP 5371 (1989+) appears bit earlier than the article suggests "2000-ies".
I keep learning , Thanks
Robert

[rdl]

This guy does a review of some chinese frequency counter thing. Seems to perform very well. Not sure when I'd need out outside of RF.

There's another thread here somewhere about those frequency counter modules. That one doesn't look bad at all. I guess I'll have to buy one. I haven't been able to find a decent deal on ebay for something used that I like (don't want to spend more than $150 or so).

Also, thanks for that youtube link. I hadn't seen that guy's videos before. I've watched a few more of them and they are nicely done and very informative.

[HarryDoPECC]

Some additional information for this necro-thread.

Reciprocal method and interpolation go back further - HP introduced both of these with the 5360A of 1970.  Well worth a read of the May69 and Dec70 HP Journals for a detailed description.  HP5323A and 5326A offered same technologies for those who did not have the entry price of well over 50k (todays $) for a 5360A.

All of these were pre-microprocessor, the maths was implemented in discrete TTL and the digital front end had bleeding edge ECL for maximim speed.  I'm wrestling with the 500+ ICs of a 5360A whenever I feel like a challenge....