Of course not.
There are two or more conversation going on right now which add to the confusion.
One is about canceling distortion in a gen and the other canceling distortion as means of analysis.
I can't tell which reply is referring to which conversation some times.
Last edited:
Hi David,
Thanks, but I wouldn't call myself an expert. I just know a little about them.
Some oscillator units have separate power connections for the heater circuit. Since they involve heavier currents (and hopefully proportional control), it makes sense to power them from another unregulated source. You can even feed the heater and temperature control circuits from a slightly higher voltage, which will reduce the average current to a small degree. The frequency and crystal amplifier sections are far more sensitive to power supply variations, even though they probably contain a regulator inside as part of the system. Therefore it makes a great deal of sense to feed them separately and deliver exactly what the spec sheet says in the way of voltage. The cleaner, the better.
So we have a high current supply that has high demands placed on it when the system is started from cold, and another that has maybe a couple hundred mA at the most demanded from it in a relatively constant basis. If you lose power, you need to keep the crystal at it's set temperature (each crystal is different depending on their temperature vs frequency inflection point). But the GPS receiver and frequency control circuit really should remain powered up. They are both important, but you would like to continue to track the satellites if possible. The GPS antenna also requires current (from the GPS receiver circuit). It would help if the heat loss from the oven is minimized, but keep in mind that heat flow out is required in order to pull the temperature down if it rises too high (controller overshoot for example).
Anyway, this says to me that the controller can run off a gel cell all the time while the cell is recharged constantly, like an on-line UPS system. The heaters would require larger gel cells and can run off a supply that changes over to battery through voltage drop. That wouldn't affect the heaters that much, so you end up with an unregulated and simpler power supply. Simple is good.
That idea would create a robust GPS frequency standard system. The larger labs use that idea but create three or more copies of the same equipment. They also receive frequency variation reports for varying atmospheric conditions. That corrects for differences in each satellite's height above the earth and other factors needed for extreme accuracy. Without that information, you can still achieve accuracy with errors in the range of 10 exp -13. Substantially better than the high accuracy oscillators spec of errors down in the 10 exp -9 range. Being continuously corrected for frequency means you don't need the classic calibrations to be done every couple years, or yearly for older equipment.
If all goes well, I will have three systems running off the same antenna. I might mount a backup antenna inside. There is a nifty coupler é splitter that Agilent made that allows four GPS receivers to share the one antenna. The antenna is powered by the first receiver. Most of this equipment uses N connectors. Probably 75 R because it is an antenna, but I haven`t checked. My cable to the antenna is rated 75 R. If that`s wrong I'll have a SWR of 1:1.5 which shouldn`t hurt me. I installed the cables in 2001 and cannot remember the type of coax I used. It was intended for satellite TV and worked well for that.
-Chris
Thanks, but I wouldn't call myself an expert. I just know a little about them.
Some oscillator units have separate power connections for the heater circuit. Since they involve heavier currents (and hopefully proportional control), it makes sense to power them from another unregulated source. You can even feed the heater and temperature control circuits from a slightly higher voltage, which will reduce the average current to a small degree. The frequency and crystal amplifier sections are far more sensitive to power supply variations, even though they probably contain a regulator inside as part of the system. Therefore it makes a great deal of sense to feed them separately and deliver exactly what the spec sheet says in the way of voltage. The cleaner, the better.
So we have a high current supply that has high demands placed on it when the system is started from cold, and another that has maybe a couple hundred mA at the most demanded from it in a relatively constant basis. If you lose power, you need to keep the crystal at it's set temperature (each crystal is different depending on their temperature vs frequency inflection point). But the GPS receiver and frequency control circuit really should remain powered up. They are both important, but you would like to continue to track the satellites if possible. The GPS antenna also requires current (from the GPS receiver circuit). It would help if the heat loss from the oven is minimized, but keep in mind that heat flow out is required in order to pull the temperature down if it rises too high (controller overshoot for example).
Anyway, this says to me that the controller can run off a gel cell all the time while the cell is recharged constantly, like an on-line UPS system. The heaters would require larger gel cells and can run off a supply that changes over to battery through voltage drop. That wouldn't affect the heaters that much, so you end up with an unregulated and simpler power supply. Simple is good.
That idea would create a robust GPS frequency standard system. The larger labs use that idea but create three or more copies of the same equipment. They also receive frequency variation reports for varying atmospheric conditions. That corrects for differences in each satellite's height above the earth and other factors needed for extreme accuracy. Without that information, you can still achieve accuracy with errors in the range of 10 exp -13. Substantially better than the high accuracy oscillators spec of errors down in the 10 exp -9 range. Being continuously corrected for frequency means you don't need the classic calibrations to be done every couple years, or yearly for older equipment.
If all goes well, I will have three systems running off the same antenna. I might mount a backup antenna inside. There is a nifty coupler é splitter that Agilent made that allows four GPS receivers to share the one antenna. The antenna is powered by the first receiver. Most of this equipment uses N connectors. Probably 75 R because it is an antenna, but I haven`t checked. My cable to the antenna is rated 75 R. If that`s wrong I'll have a SWR of 1:1.5 which shouldn`t hurt me. I installed the cables in 2001 and cannot remember the type of coax I used. It was intended for satellite TV and worked well for that.
-Chris
A mismatch in impedance from 75ohm to say 50ohm just mean greater attenuation and the micro reflections will be larger. You will have larger standing waves within the system but probably not large enough to cause serious problems. It just make things a bit lumpy from a broadband perspective if that make any sense to you.
I suppose you could reject some of that fund 0dB level and still be able to lock??
Bob's DM needs a minimum gain of 20dB to function well.... i had one. But what to do when measuring buffers or less than a X10 amp. Or passive parts?
THx-RNMarsh
Last edited:
Scott,
Let's compare apples to apples. To measure distortion a bridge comparing source to output is a great method. I have used it often and for even more than just amplifier testing.
To determine the distortion of an oscillator a bridge has no reference. The Stanford Research Systems SR850 will resolve to -100 or better and uses a synthesized clock reference. I have found having the oscillator and lock in amplifier fed from the same clock improves things.
But if you want to add a bridge between the original sine wave and the distortion compensated one I suspect you could measure the vibrations from the angels dancing on the head of a pin in the adjoining room. 🙂
For my testing I often use an FIR generated pair of sine waves that are crystal controlled and then double resonator filtered. This is better than any device I have ever tested this way. But the error signal for some tests is still so low long time averaging is required.
But that is not what I am currently concerned with exploring. I did build some cute equipment along those lines and it pointed to other issues I still need to play with.
My conclusion is that currently it is possible to get total harmonic distortion low enough that it is not the limiting issue for reproduction or reinforcement of audio. (That does not mean that all equipment meets those limits.)
Doesnt seem to make any difference with test result under these conditions. I also had a freq counter and an ac vm on the Tport and saw no increase in distortion.
With the specific equipment I used. [CYA statement]
-RNM
Last edited:
Hi Richard, that is true, but that is just a design choice of an internal attenuation path so that I could measure power amplifiers. If I wanted to, I could have added a switchable attenuator, perhaps in a simple 1, 2, 5 sequence. One could also arrange it so that one could put one of Jan's auto-scalers in front of it.
Cheers,
Bob
OK. But too late. I dont usually modify hardware or software and try to find what I can that works for me right out of the box.
A LP filter of high atten rate just after the generator output seems to do the job pretty well. Can improve gen of simpler and higher distortion into a clean pure output for testing. --- as an alternative to more complex cancellation methods. Seems worth a try at least. But cancellation seems -on paper- to be a nice solution also.
After the ultra clean gen is done and any residual is removed.... then we have a variable freq source which can be used to test anything for the rest of our life.... and not just audio amplifiers. No more playing catch-up between DUT and test gens and analyzers. Well, at least to -160dBv. I can live with that.
[I have to go find a water leak up on the roof somewhere before it rains again.....]
THx-RNMarsh
Last edited:
The LP filter after the oscillator is a fairly obvious thing to do, but it is only practical if it is tuned in step with the oscillator. Not that that cannot be done, but it can be a PIA. That LP filter might actually be an SVF that is tuned with another set of the same tuning elements ganged with the tuning elements as the oscillator.
The main reason for the DM is not to get the equivalent of an ultra-low audio oscillator source (but it is a nice side benefit), but rather to increase the dynamic range of the THD analyzer or the spectrum analyzer that follows it.
Analyzer dynamic range and its residual distortion is often the long pole in the tent.
Cheers,
Bob
yes, and so should be tried. Simple test i did showed it works very well to eliminate harmonics. I used a miniDSP set to 48db/oct at just above osc freqs and cleaned up a modest little gen well. but is too noisy for my goals. Just proof of concept in how effective it can be.
So, sync'ing it with the fundamental gen freq is all that would needed to be done.... and there are several ways to do that. Maybe some others here can brain-storm some additional ways to do it. ??
THx-RNMarsh
I thought of this when looking over the Boonton schematics and how easy it would be to have them track
with computer control and both having the same tuning elements. They could even have similar PCB layouts
for the SV osc, LP, and notch. Did you consider gain error introduced by the LP filter? Not sure what I'd do
about that.
Did you have a look at the osc gain control in the Boonton by any chance?
Last edited:
It would probably be simpler to notch the remaining distortion out.
If the Q were much lower than what we normally use it will include both harmonics.
On second thought a passive Twin T tuned to the second H would get both Harmonic.
A high order analog low pass would be very noisy.
I like Bob's idea of a synchronously tuned state variable filter. Not cheap though.
Last edited:
I like simple.
After all the investment i have made in this quest.... cheap does not come to mind as an over-riding concern.
-Richard
After all the investment i have made in this quest.... cheap does not come to mind as an over-riding concern.
-Richard
Last edited:
We designed voltage controlled, tracking, state variable filters in graduate school.
The HP and LP outputs have equal amplitude when on the fundamental which
makes it easy to use an integrator loop to track. CA3080s were used to provide
voltage control. Going to guess that the distortion is not low enough for this type
of application with the voltage control.
The HP and LP outputs have equal amplitude when on the fundamental which
makes it easy to use an integrator loop to track. CA3080s were used to provide
voltage control. Going to guess that the distortion is not low enough for this type
of application with the voltage control.
Richard, So if I read you correctly, then for your 1000Hz test signal
ran through a miniDSP set the low pass at what freq w/the -48dB/octive
slope? Are we talking 1050Hz or 1100Hz or 1001Hz?
I may have a hardware/software solution that might be worth
looking into. I plan to use it for my speaker cabinet upgrade
project for tri-stereo amping the speakers with active x-over.
I have an extra output channel for L, R reserved for a Sub Woofer.
I don't want to say anything else here as I'll never find another affordable
source for this gear again.
ran through a miniDSP set the low pass at what freq w/the -48dB/octive
slope? Are we talking 1050Hz or 1100Hz or 1001Hz?
I may have a hardware/software solution that might be worth
looking into. I plan to use it for my speaker cabinet upgrade
project for tri-stereo amping the speakers with active x-over.
I have an extra output channel for L, R reserved for a Sub Woofer.
I don't want to say anything else here as I'll never find another affordable
source for this gear again.
Both active filters and passive filters will have distortions that limit the eventual performance. A synchronously tuned SV filter might be a simple trick, especially if you already have the SV oscillator PCB. To stabilize the output wrap the AGC from the filter output. If you have enough resolution with the stepped components (fixed R's and C's) a PLL on the oscillator would probably keep the output inside the working range of the SV filter so it would not need fine tuning. That cheap SVO board from China would be a good starting point.
You could use another pair for an analyser with two stages of notch.
At this point the output amp becomes the limitation. An output amp with enough power to drive 50 Ohms to even 3V with distortion products below -150 dB is a tall order.
Another option might be a cap low pass filter on the output inside the AGC again. it would need a lot of caps to encompass the complete range.
ir a handful of fixed frequency oscillators with passive filters on the output. It may be possible to make LC filters but the L's would be quite custom (and BIG).
You could use another pair for an analyser with two stages of notch.
At this point the output amp becomes the limitation. An output amp with enough power to drive 50 Ohms to even 3V with distortion products below -150 dB is a tall order.
Another option might be a cap low pass filter on the output inside the AGC again. it would need a lot of caps to encompass the complete range.
ir a handful of fixed frequency oscillators with passive filters on the output. It may be possible to make LC filters but the L's would be quite custom (and BIG).
may not be falling of a log easy, but for a composite op amp circuit it looks quite possible with today's better op amps
like 400 mA, 100 MHz CFA DSL drivers in multi Watt Pdiss PowerPad packages
in inverting gain in particular it is just compensating the loop(s) with a low noise input op amp wrapping its gain around a CFA designed to drive DSL a few W into 25 Ohms primary of the line matching xfmers
TPA6120 CFA (rebranded THS6015) was my choice, I've paralleled pairs' outputs, biased them into push-pull Class A
with a OPA627 as input op amp wrapping the paralleled TPA6120, it measured > 120 loop gain to 30 kHz in hardware (well there was one more op amp adding loop gain too - that makes the compensation much more interesting)
some info a few pages around:
http://www.diyaudio.com/forums/equi...n-audio-range-oscillator-368.html#post3938612
