Why would you need more than 4 place resolution?
Surely your stray inductance will be vastly more than that?
Or was this just a "heads up" for other applications?
Best wishes
David
Last edited:
Hi Phil,
That is highly variable. It depends on the timing source and whether or not there is a precision oscillator in the unit. It also depends on the brand. For example, my HP 5335A's (I have two), they have never been out of tolerance. They also both use an ovenized oscillator which maintains a constant temperature for the crystal, which is also supplied with regulated voltages. In my case, all my equipment that will accept an external reference now is supplied with a GPS disciplined clock (10 MHz) supplied through a proper distribution amplifier (HP 5087A in my case). I would expect a Racal Dana counter to also remain within tolerance and all the crosschecks I performed on the calibration lab's assets showed it to be in tolerance. I would be comfortable using these in business with a 5 year calibration interval, and on the GPS system, I would simply perform crosschecks every year as I have another GPS receiver going on line soon (I hope) that can be used for sanity checks. I also will have another ovenized HP oscillator running independently that will be used for sanity checks.
Cheaper counters that use a crystal (not ovenized) would have to be left on so that internal temperatures (your big enemy is temperature shifts) are as stable as they can be. They can be completely unreliable, calibration every three months, to models that us a good ovenized oscillator that should be good for every 2 years or longer depending on their calibration history.
There are ovenized oscillators that can be purchased new, and a newer type called a compensated oscillator. These sense the ambient temperature and electronically adjust the frequency comparing to a lookup table within the unit. Of course the quality of these will determine what your calibration cycle should be.
For the careful hobbyist who needs reasonable precision and confidence, I would recommend the following. Buy an HP frequency counter used, the most recent and highest quality you can afford. Have it calibrated once you get it and if you can afford it, get the calibration with full data. There is a reason for this, one of which being that the lab that does the work actually performs it properly. Use a lab that is national with more than one lab. That means they keep their standards to the levels they say they do. Now, with your new counter and certification, plug your counter in and use it with confidence for years and years. Unless something breaks, the chances are it will remain in tolerance for a decade or more. Now, you will notice the oscillator and oven remain powered on even though the instrument is turned off. It can take up to a week for a really good frequency source to come back into tolerance after being turned off. Putting it on a small UPS wouldn't be overly cautious.
Okay, so the above situation should give you a frequency source and counter good to the 10 exp(-9) range in stability and accuracy. You can update a cheaper counter with a compensated oscillator to be much more accurate (it would need calibration as soon as you begin using it), but not an ovenized oscillator. The reason is that the oven takes a few amps until it has reached operating temperature. It needs it's own heater and oscillator power supplies.
Another option is to get your hand on a new or used ovenized oscillator and put it in a case with power supply. Get it calibrated before use, then you can plug it into almost any quality frequency counter (or frequency generator, spectrum or network analyzer) and instantly upgrade it's performance. With a distribution amplifier, this one frequency source will upgrade every instrument you own that will accept an external frequency reference. Normally that is 10 MHz. The only item that might need calibration would be your frequency source in that case.
The inexpensive counters are pointless for any close work unless you upgrade their oscillator. The other things that matter will be the quality of gate and signal processing to reduce jitter (sound familiar?). Your frequency counter should trigger reliably in order to be accurate. Some also have attenuators and filtering to help reduce noise on your signal.
Sorry for the long post, but it couldn't be answered in a couple lines. There are still more factors to consider, but the upshoot really is, buy an HP / Agilent / Keysight frequency counter. Make sure it is in tolerance before you start using it, then try to keep the room temperature reasonably constant and the AC power supplied continuously. If you buy an oscillator, you have aging issues to deal with. That means that a new oscillator should be calibrated more often starting out, and once it has proved itself to be stable, you can almost forget about it. So buying used can actually be a benefit.
-Chris
That is highly variable. It depends on the timing source and whether or not there is a precision oscillator in the unit. It also depends on the brand. For example, my HP 5335A's (I have two), they have never been out of tolerance. They also both use an ovenized oscillator which maintains a constant temperature for the crystal, which is also supplied with regulated voltages. In my case, all my equipment that will accept an external reference now is supplied with a GPS disciplined clock (10 MHz) supplied through a proper distribution amplifier (HP 5087A in my case). I would expect a Racal Dana counter to also remain within tolerance and all the crosschecks I performed on the calibration lab's assets showed it to be in tolerance. I would be comfortable using these in business with a 5 year calibration interval, and on the GPS system, I would simply perform crosschecks every year as I have another GPS receiver going on line soon (I hope) that can be used for sanity checks. I also will have another ovenized HP oscillator running independently that will be used for sanity checks.
Cheaper counters that use a crystal (not ovenized) would have to be left on so that internal temperatures (your big enemy is temperature shifts) are as stable as they can be. They can be completely unreliable, calibration every three months, to models that us a good ovenized oscillator that should be good for every 2 years or longer depending on their calibration history.
There are ovenized oscillators that can be purchased new, and a newer type called a compensated oscillator. These sense the ambient temperature and electronically adjust the frequency comparing to a lookup table within the unit. Of course the quality of these will determine what your calibration cycle should be.
For the careful hobbyist who needs reasonable precision and confidence, I would recommend the following. Buy an HP frequency counter used, the most recent and highest quality you can afford. Have it calibrated once you get it and if you can afford it, get the calibration with full data. There is a reason for this, one of which being that the lab that does the work actually performs it properly. Use a lab that is national with more than one lab. That means they keep their standards to the levels they say they do. Now, with your new counter and certification, plug your counter in and use it with confidence for years and years. Unless something breaks, the chances are it will remain in tolerance for a decade or more. Now, you will notice the oscillator and oven remain powered on even though the instrument is turned off. It can take up to a week for a really good frequency source to come back into tolerance after being turned off. Putting it on a small UPS wouldn't be overly cautious.
Okay, so the above situation should give you a frequency source and counter good to the 10 exp(-9) range in stability and accuracy. You can update a cheaper counter with a compensated oscillator to be much more accurate (it would need calibration as soon as you begin using it), but not an ovenized oscillator. The reason is that the oven takes a few amps until it has reached operating temperature. It needs it's own heater and oscillator power supplies.
Another option is to get your hand on a new or used ovenized oscillator and put it in a case with power supply. Get it calibrated before use, then you can plug it into almost any quality frequency counter (or frequency generator, spectrum or network analyzer) and instantly upgrade it's performance. With a distribution amplifier, this one frequency source will upgrade every instrument you own that will accept an external frequency reference. Normally that is 10 MHz. The only item that might need calibration would be your frequency source in that case.
The inexpensive counters are pointless for any close work unless you upgrade their oscillator. The other things that matter will be the quality of gate and signal processing to reduce jitter (sound familiar?). Your frequency counter should trigger reliably in order to be accurate. Some also have attenuators and filtering to help reduce noise on your signal.
Sorry for the long post, but it couldn't be answered in a couple lines. There are still more factors to consider, but the upshoot really is, buy an HP / Agilent / Keysight frequency counter. Make sure it is in tolerance before you start using it, then try to keep the room temperature reasonably constant and the AC power supplied continuously. If you buy an oscillator, you have aging issues to deal with. That means that a new oscillator should be calibrated more often starting out, and once it has proved itself to be stable, you can almost forget about it. So buying used can actually be a benefit.
-Chris
For $200 or less you can get a GPSDO (GPS Disiplined Oscillator) 10 MHz reference receiver from eBay. No calibration ever needed. I have compared two receivers and there may be a slow phase drift over time, like 5 minutes for 90 degrees at 10 MHz. That's really small. Getting an oveninzed oscillator calibrated is more expensive than just buying a GPSDO.
That receiver can be used as a reference in many counters. I have an HP 5316A and an HP 5370A w/ Beaglebone Black upgrade. What I have learned re: audio is that most internal clocks meet the SPDIF standard but aren't right on frequency. Some PC sources are way off and for audio frequency the extra digits will drive you nuts. Also you need a reciprocal counter for audio, otherwise it takes too long for a reading.
The 5370 can measure down to 20 pS which is amazing but not good enough for the best audio digital chains so it sits on a cart.
That receiver can be used as a reference in many counters. I have an HP 5316A and an HP 5370A w/ Beaglebone Black upgrade. What I have learned re: audio is that most internal clocks meet the SPDIF standard but aren't right on frequency. Some PC sources are way off and for audio frequency the extra digits will drive you nuts. Also you need a reciprocal counter for audio, otherwise it takes too long for a reading.
The 5370 can measure down to 20 pS which is amazing but not good enough for the best audio digital chains so it sits on a cart.
So far the OP has indicated " to use for audio experiments"; "measuring small signal audio"; "measuring T/S parameters, for viewing waveform distortion with a scope, for playing around with filters, that sort of thing
"; " to check for consistency and keep a watch on equipment faults and user error...".To my mind that doesn't sound like a 10 digit accuracy frequency counter that needs to stay on 24/7, or a new user needing a dedicated hardware frequency synthesizer, when they have a scope but find it a bit difficult/tiring to check on the frequency of a waveform.
There are diy and cheap interfaces around for soundcards, as well as commercial USB soundards with 1Meg input impedance and input gain pots that connect directly to low signal level audio projects. The software/hardware may not be calibrated for 10 digit frequency accuracy, but I'm sure it would be better than 1%, and include a whole bevy of standard and specialised signals such as various IM modes, and all the noise colours, with bandwidths out to at least 48kHz, and likely 96kHz, and sine distortion well below 0.01%.
Hi David, I recommend you work through an example. It's illuminating.
Assume C1=9.49E-9 and L1=1.70E-8, solve for f1
Assume C2=9.49E-9 and L2=1.71E-8, solve for f2
How many display digits are needed, to view the difference between f1 and f2?
How many display digits are needed, to accurately calculate L2 given TruncatedDisplay(f2), and to accurately calculate L1 given TruncatedDisplay(f1) ?
_
Last edited:
I wouldn't even bother with a handheld frequency counter... We live in a day and age where a two channel oscilloscope is the better tool to have laying about. You could use the FFT function for frequency measurement. The scopes trigger modes, cursors and memory comes in handy for diagnosing and/or understanding all sorts of timing issues for many other things..
Hi Phil,
Thanks!
Hi jrdmedford,
The problem with using a spectrum / network analyser of 'scope FFT function is that the reported peak is rounded to the closest bin. Some will allow an exact measurement with some extra keystrokes. A frequency counter is the correct tool for this and by far the easiest to use.
Hi Keith,
The frequency counter built into the 34401A would be my first choice, the 5385A my next choice based on update rate for low frequencies. The HP 34401A gets my first choice because it does a huge amount of things for you along with a great frequency counter. Buying one of these would be well advised. Otherwise the HP 5385A represents the best value and reliability.
The GPSDO that Demian suggested is an excellent choice too if you have a frequency counter that has a ref in jack. Keep in mind that short term jitter is worse with the GPSDO due to constant correction. If you need very low jitter, put it in holdover mode for the measurement, then restore the steering of the oscillator.
I haven't seen any for less than $200 lately. I have a T-Bolt from the Timenuts group buy which is working well. I have a Symmetricom (?sp) that I'll be putting into service soon when the panel is ready for it. Too bad it has 5 MHz and 15 MHz outputs instead of a 10 MHz output with low jitter. I'm pretty sure it also has a 1 pps output that I'll be using for another locked oscillator at 10 MHz. This is a hobby all onto itself.
Hi Demian,
I also have a 5370A I use occasionally. It's a pretty capable instrument. I use it for checking oscillator health in any number of things.
-Chris
Thanks!
Hi jrdmedford,
The problem with using a spectrum / network analyser of 'scope FFT function is that the reported peak is rounded to the closest bin. Some will allow an exact measurement with some extra keystrokes. A frequency counter is the correct tool for this and by far the easiest to use.
Hi Keith,
The frequency counter built into the 34401A would be my first choice, the 5385A my next choice based on update rate for low frequencies. The HP 34401A gets my first choice because it does a huge amount of things for you along with a great frequency counter. Buying one of these would be well advised. Otherwise the HP 5385A represents the best value and reliability.
The GPSDO that Demian suggested is an excellent choice too if you have a frequency counter that has a ref in jack. Keep in mind that short term jitter is worse with the GPSDO due to constant correction. If you need very low jitter, put it in holdover mode for the measurement, then restore the steering of the oscillator.
I haven't seen any for less than $200 lately. I have a T-Bolt from the Timenuts group buy which is working well. I have a Symmetricom (?sp) that I'll be putting into service soon when the panel is ready for it. Too bad it has 5 MHz and 15 MHz outputs instead of a 10 MHz output with low jitter. I'm pretty sure it also has a 1 pps output that I'll be using for another locked oscillator at 10 MHz. This is a hobby all onto itself.
Hi Demian,
I also have a 5370A I use occasionally. It's a pretty capable instrument. I use it for checking oscillator health in any number of things.
-Chris
My primary GPSDO (Arbiter Systems) uses a Wenzel 5 MHz streamline and is pretty free of jitter. Optimum internal firmware should exploit the known qualities of the OXCO and not overcorrect or otherwise degrade it. The Thunderbolts are quite tunable and can be connected to other oscillators. Tuning them for optimal performance can be as consuming as high end audio.
However for optimizing speakers this may be gross overkill unless you keep the drivers in temperature and humidity controlled environments.
The 5370 is quite a beast. Not really good for frequency but about as good as you will get for time intervals. The processor upgrade is nice and provides network access.
However for optimizing speakers this may be gross overkill unless you keep the drivers in temperature and humidity controlled environments.
The 5370 is quite a beast. Not really good for frequency but about as good as you will get for time intervals. The processor upgrade is nice and provides network access.
Hi Demian,
I got the 5370A a short time ago and haven't really got into it. I have the 50R input modules, and a 1 Meg module with 10:1 probe # 54003A. The kit looks like it was never used. Can you PM me some info on the upgrade?
My other GPSDO is an HP# Z3811A and Z3812A combination. Brand new sealed in a box. It is marked Lucent - which would be Bell Labs to the best of my knowledge from my phone work. I have a rack mount panel coming for it to also combine the T-bolt and a free running OXCO oscillator (HP E1938). I haven't had a chance to tune the T-bolt yet as I'm waiting for something to mount it to. It's just sitting behind the distribution amplifier now. It gets pretty dusty there.
-Chris
Edit: Yes, overkill for sure. Mine is for some RF work, and it's nice to have all the instruments read the same frequency exactly. It's also nice not to have to worry about frequency accuracy.
I got the 5370A a short time ago and haven't really got into it. I have the 50R input modules, and a 1 Meg module with 10:1 probe # 54003A. The kit looks like it was never used. Can you PM me some info on the upgrade?
My other GPSDO is an HP# Z3811A and Z3812A combination. Brand new sealed in a box. It is marked Lucent - which would be Bell Labs to the best of my knowledge from my phone work. I have a rack mount panel coming for it to also combine the T-bolt and a free running OXCO oscillator (HP E1938). I haven't had a chance to tune the T-bolt yet as I'm waiting for something to mount it to. It's just sitting behind the distribution amplifier now. It gets pretty dusty there.
-Chris
Edit: Yes, overkill for sure. Mine is for some RF work, and it's nice to have all the instruments read the same frequency exactly. It's also nice not to have to worry about frequency accuracy.
It's anything but illuminating: why on Earth would you be interested in accurately measuring some negligibly small difference between two small parasitic inductances?
Hi Marcel,
Looking at the Ebay prices I saw yesterday, a really good HP 5385A counter is $250. Much less accurate counters are sitting at $100 and change and the junk is less. May as well go for the one good for decades. It is common for old techs to have equipment they bought three or more decades ago. I have some that old and it still works perfectly.
The right answer can be the difference between your project working and not. Wouldn't you want the right information working on a project?
-Chris
The difference is 0.01 units, and that can be important in tuned circuits. More importantly, it is all about knowing you have the right answer and the less expensive counters can be miles off. The example is only three digit and one LSD difference. This could be +/- one count either way for starters and that represents a huge error. So you would want to have minimum 4 digit accuracy and preferably more. Since this is easy to do, why not get the right answer?
Looking at the Ebay prices I saw yesterday, a really good HP 5385A counter is $250. Much less accurate counters are sitting at $100 and change and the junk is less. May as well go for the one good for decades. It is common for old techs to have equipment they bought three or more decades ago. I have some that old and it still works perfectly.
The right answer can be the difference between your project working and not. Wouldn't you want the right information working on a project?
-Chris
I can understand the importance of 100 pH in some GHz-range tuned circuit, but in post 14, Mark refers to a thread about the parasitic inductance of electrolytic capacitors. If it really matters to you whether an electrolytic capacitor has an ESL of 17 nH or 17.1 nH, there must be something seriously wrong with your design.
This here is my Lucent pair. The Z3811A is modded for 10 MHz and TTL 1PPS out.
The modification is described here:
< http://www.hoffmann-hochfrequenz.de/downloads/DoubDist.pdf >
I have also bought quite a lot of GPS-less Z3812A for their MTI-260 ovens
when the offer was right on ebay. They have been built by Agilent for Lucent
and stocked as spare parts. I had not much luck with crystal ovens from China,
these came as originally shrink-wrapped from FLA.
I'd like to phaselock 16 of them very slowly and then powercombine them
for averaging phase noise. BVAs are no longer available, so there is
some need for a replacement.
I have made a carrier board for that purpose that can house an MTI-260,
10811A or MV-89.
The 10811A would be the one from my 5370A, which is just an array of
intermittent contacts. Hopeless. Bought the SR620 instead.
< dsc_472 | Gerhard Hoffmann | Flickr >
< dsc_0471 | OCXO carrier board. Hosts an MTI-260, Morion MV89… | Flickr >
Cheers,
Gerhard
(That has no application whatsoever for measuring TS parameters or audio filters.)
The modification is described here:
< http://www.hoffmann-hochfrequenz.de/downloads/DoubDist.pdf >
I have also bought quite a lot of GPS-less Z3812A for their MTI-260 ovens
when the offer was right on ebay. They have been built by Agilent for Lucent
and stocked as spare parts. I had not much luck with crystal ovens from China,
these came as originally shrink-wrapped from FLA.
I'd like to phaselock 16 of them very slowly and then powercombine them
for averaging phase noise. BVAs are no longer available, so there is
some need for a replacement.
I have made a carrier board for that purpose that can house an MTI-260,
10811A or MV-89.
The 10811A would be the one from my 5370A, which is just an array of
intermittent contacts. Hopeless. Bought the SR620 instead.
< dsc_472 | Gerhard Hoffmann | Flickr >
< dsc_0471 | OCXO carrier board. Hosts an MTI-260, Morion MV89… | Flickr >
Cheers,
Gerhard
(That has no application whatsoever for measuring TS parameters or audio filters.)
Attachments
Last edited:
Hi Gerhard,
Many thanks for the links. That Lucent is the exact thing I have. Looks like I will be modifying it at some point in time.
I have the same ovens in mine. They seem to be okay. I also have an HP 10811A that I can use as a stand alone oscillator for sanity checks. The E1938A is a newer oscillator. I have a pair, one has a dead oven that I'll try to fix.
-Chris
Many thanks for the links. That Lucent is the exact thing I have. Looks like I will be modifying it at some point in time.
I have the same ovens in mine. They seem to be okay. I also have an HP 10811A that I can use as a stand alone oscillator for sanity checks. The E1938A is a newer oscillator. I have a pair, one has a dead oven that I'll try to fix.
-Chris
WORK!? That's why I asked you the question, so I wouldn't have to
Ok, I have F1 = 78.730 4 MHz
and F2 = 78.499 9 MHz ...unless I had a slip on my trusty HP 11c
So to 4 places 78.73 and 78.50 or 78.49, dependent on luck on LSB.
That's 23 or 24 counts, better than 5% uncertainty on a difference of 0.1 nH.
Now 0.1 nH is the extra inductance of about 0.1 mm or wire (0.004 of an inch for the metrically handicapped).
Even with only 4 places the 5% uncertainty of the difference corresponds to uncertainty in the inductance of a few microns of wire.
I just can't see it matters, I understand that we want as accurate an answer as possible but it looks to me as if the uncertainty is dominated by other issues.
And even more so with 5 place resolution.
This is, of course, independent of the question of whether the 0.1 nH difference even matters in the first place-
I'm happy to just look at the principle because your comments are educational to think thru, I just realized it's exactly what the instructors tried to make me understand in first year physics prac. work - "how do small variations in measurements show up as variation in the results?".
For LC resonance the frequency is a square root function of the inductance so for ~1% result resolution you need ~0.5% measurement resolution, it's not equal.
In your example it's about a 0.6% variation in inductance that reflects as a 0.3% frequency variation.
This was my rule of thumb why 4 places would be OK, it's better than 0.1%.
So thanks, and it reminds me of another point of interest.
I looked at your capacitor results and they seem a bit unusual.
IIRC the capacitor manufacturers say the stray ESL is close to that of a piece of wire the distance between the terminals.
Your results have the inductance fall as the distance increases, any comments?
Best wishes
David
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.