I typically use 5651 gas reference tubes in my voltage regulators because I like the relative lack of thermal voltage coefficients and they are pretty consistent voltage wise. I have not been able to find any discussions pertaining to their ac small signal behavior. (Analog to dynamic impedance in a zener for example.)
I have searched vainly in RD4, ARRL handbooks, and online for a discussion on the negative resistance characteristic and the internal or dynamic impedance characteristic of these tubes. The static resistance is easy enough to figure out, but I am interested in what happens when a small ac component is superimposed.. The reason for this is that I use them in the cathode of my cascode error amplifier design, and because they cannot be bypassed by significant amounts of capacitance I am wondering what the effective impedance over the audio band might be - as this has a direct bearing on the open loop gains of my regulators and hence regulator performance.
Based on some recent measurements and some stability issues I have in some newer designs I believe the internal impedance of the gas reference is quite significant.
The only thing I can think to do is come up with a jig to measure typical performance over the frequency range of interest at the design operating current.
Any thoughts?
I have searched vainly in RD4, ARRL handbooks, and online for a discussion on the negative resistance characteristic and the internal or dynamic impedance characteristic of these tubes. The static resistance is easy enough to figure out, but I am interested in what happens when a small ac component is superimposed.. The reason for this is that I use them in the cathode of my cascode error amplifier design, and because they cannot be bypassed by significant amounts of capacitance I am wondering what the effective impedance over the audio band might be - as this has a direct bearing on the open loop gains of my regulators and hence regulator performance.
Based on some recent measurements and some stability issues I have in some newer designs I believe the internal impedance of the gas reference is quite significant.
The only thing I can think to do is come up with a jig to measure typical performance over the frequency range of interest at the design operating current.
Any thoughts?
Hi Kevin,
Good question. I've measured the larger VR tubes in the past but have not measured the little 5651. With some digging around in the stash I was able to fine one Raytheon 5651WA.
I spent a lot of time last week designing a new version of the power supply tester and a ripple injector for testing power supplies. This looked like a good task for the power supply tester.
First, the test circuit.
As the power supply tester needs a minimum of 10ma to run properly, and is a pita to set at low currents I set it for 18ma. The CCS feeding the 5651 is set to 20.5ma. This sets the current through the 5651 to 2.5ma as measured by the Fluke 8020A DMM.
Here is a shot of the broad band noise generated by the 5651. You can see the frequency distribution of the noise and at the lower left is the total noise level, -47.54 re 0dBV. Oops, just noticed that my comment referring to 0dB is incorrect. The charts show 0dBV = .775V which is incorrect. They should say 0dBV = 1V.
While setting the AC current level used for testing the impedance vs frequency the distortion of the 5651 was visible.
Here is the distortion at 1Khz of the 5651. DC current was 2.54ma and the AC current was .35ma AC RMS.
Last up is the dynamic impedance of the 5651. The chart says it all. The impedance in dB is noted at the top of the chart.
Hope this fits your needs.
Gary
Good question. I've measured the larger VR tubes in the past but have not measured the little 5651. With some digging around in the stash I was able to fine one Raytheon 5651WA.
I spent a lot of time last week designing a new version of the power supply tester and a ripple injector for testing power supplies. This looked like a good task for the power supply tester.
First, the test circuit.
An externally hosted image should be here but it was not working when we last tested it.
As the power supply tester needs a minimum of 10ma to run properly, and is a pita to set at low currents I set it for 18ma. The CCS feeding the 5651 is set to 20.5ma. This sets the current through the 5651 to 2.5ma as measured by the Fluke 8020A DMM.
Here is a shot of the broad band noise generated by the 5651. You can see the frequency distribution of the noise and at the lower left is the total noise level, -47.54 re 0dBV. Oops, just noticed that my comment referring to 0dB is incorrect. The charts show 0dBV = .775V which is incorrect. They should say 0dBV = 1V.
An externally hosted image should be here but it was not working when we last tested it.
While setting the AC current level used for testing the impedance vs frequency the distortion of the 5651 was visible.
Here is the distortion at 1Khz of the 5651. DC current was 2.54ma and the AC current was .35ma AC RMS.
An externally hosted image should be here but it was not working when we last tested it.
Last up is the dynamic impedance of the 5651. The chart says it all. The impedance in dB is noted at the top of the chart.
An externally hosted image should be here but it was not working when we last tested it.
Hope this fits your needs.
Gary
Excuse my being late... may I also congratulate Gary for sharing this info with us.
My question is what happens if a 5651 is paralleled with a capacitor? Does it help with impedance reduction at the higher frequencies? Do any nasty high-Q patterns creep in? Total noise power should also get reduced, but if significant departure from the expected white-ish noise spectrum is obtained, then perhaps it is better to leave the VR unbypassed?
Cheers.
My question is what happens if a 5651 is paralleled with a capacitor? Does it help with impedance reduction at the higher frequencies? Do any nasty high-Q patterns creep in? Total noise power should also get reduced, but if significant departure from the expected white-ish noise spectrum is obtained, then perhaps it is better to leave the VR unbypassed?
Cheers.
Agreed, excellent information on the 5651 from Gary (thanks from me also). Note that the 5651 is not a voltage regulator tube (in the classic sense) due it's very narrow operating current range (1.5ma - 3.5ma) and is listed as a voltage reference tube. It's not designed to be used as a shunt regulator but is more accurate as a voltage reference source than your typical VR tubes. You can bypass it with a low value capacitor, I've never used more than 0.01uF and I've never had any problems with them. Rated spec shows a maximum capacitance of 0.02uF.
Regards, KM
Regards, KM
Re: Re: Gas Reference Tube (5651/85A1/etc) Internal Impedance
Pretty much the only way to know the answer to my original question is to do an actual measurement which I was not equipped to do. (Since the manufacturer did not think to provide this information in the data sheet.) Gary graciously did the experiment and confirmed my empirical observations with hard data. Extremely useful for those of us who use the 5651 as a reference in the cathode of a cascode error amplifier where it degenerates the gain of the amplifier as a function of its internal impedance. (This unfortunately isn't a benefit, but something one needs to design around.)
Comparable voltage zeners incidentally can be an order of magnitude better in this regard, but generally have much stronger temperature coefficients and no pretty glow except when they burn up..
Amazing how old threads spring back to life.
G.Kleinschmidt said:
Pretty much the only way to know the answer to my original question is to do an actual measurement which I was not equipped to do. (Since the manufacturer did not think to provide this information in the data sheet.) Gary graciously did the experiment and confirmed my empirical observations with hard data. Extremely useful for those of us who use the 5651 as a reference in the cathode of a cascode error amplifier where it degenerates the gain of the amplifier as a function of its internal impedance. (This unfortunately isn't a benefit, but something one needs to design around.)
Comparable voltage zeners incidentally can be an order of magnitude better in this regard, but generally have much stronger temperature coefficients and no pretty glow except when they burn up..
Amazing how old threads spring back to life.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.