Re: Super low noise PSU for the OXCOs
First post moderation. DaveD approved it. It's confused me too.
The OXCO heaters pull over 200 mA when they turn on. The LM317 doesn't quite have the load stability to hold my desired voltage stability because of device heating. Hence the cap bank. Idea is for the current through the regulators to be held constant as the heaters turn on and off.
The 83 F 16 V cap bank was from an ebay ad for $15. I was actually looking for loose caps for custom design parameters.
4 OXCOs will take a little over 1 A on average. By keeping heat loss below total device operation dissipation, the heaters should eventually all turn off. If they are mounted in a large thermal mass with thermal grease one may presume there are no thermal differences between devices. Four on the faces of a cube with Peltier junctions and fins on the other faces. With good thermometry that should allow very precise correction.
Vctl is 1 ppm/V sensitivity and V+ is 8 ppb/V according to the datasheet. That is 10 Hz/V. I'm trying for 0.01 ppb/yr without recourse to GPS though the PPS will be recorded also with a high precision timer.
The datasheet states 0.05 ppb/s stability. LM399 guaranteed noise level is 20 uV which is 0.02 ppb. By heterodyning all combinations and analyzing the result I hope to correct for measured aging to within 1-2% over a period equal to prior history.
The DACs for this will be even more fun!
Have Fun!
Reg
On Thursday, September 7, 2023 at 03:55:48 PM CDT, Stephen Hanselman <kc4sw.io@...> wrote:
Reg,
sorry for the two close but quite the same comments. ?My browser indicated the first one had disappeared. ?Sound familiar??
S
|
Re: Super low noise PSU for the OXCOs
Reg,
sorry for the two close but quite the same comments. ?My browser indicated the first one had disappeared. ?Sound familiar??
S
|
Re: Super low noise PSU for the OXCOs
Reg,
yes, you¡¯ve gone well down the rabbit hole¡.. ?I was wondering about the cap line-up you have planned. It seems to me you are trying to build a big battery to run the OCXO¡¯s. ?My friend runs his Volvo on a bank of Super Caps so it makes sense that your supply should deliver fairly noise free DC.?
while it doesn¡¯t compare to the level you need we do RF amps and we bypass the VDD and Bias voltages a lot. We have found that a line-up of 2 (or 3) 470uF with 1uF tant, 0.1, 0.01 (2), 0.001 (2) and finally 100pF (2) on VDD plus an additional 10uF, 1uF, 0.1, etc on the gate bias. We do this for prevention of oscillations and bypassing noise before it gets to the high-gain LDMOS. Does it make sense to consider this in your application? ?The various values are used based on case resonance.
don¡¯t know if this is any help but what do you think??
steve
|
Re: Super low noise PSU for the OXCOs
Reg,
yup you¡¯ve gone way over the edge to the time/frequency nuts. ?While your filtering is impressive have you considered high frequency noise? ?When we do RF amps we put two or three 470uF electrolytics on the VDD supply. Additionally there are a pile of smaller caps, typically 1uF, 0.1 (2), 0.011(2), 0.001 (2), and finally 100pF(2) also on VDD to ground. ?Of course we are looking at the possibility of oscillations at ¡°weird¡± frequencies but given the route you are going does it make sense to consider this in your application as well? ?
I have no experience to back this up but it looks like you are very worried about very small ¡°noise¡± on the DC to the OCXO¡¯s. ?We have found that the extra caps (we use X7R for the bigger and ATC case ¡°B¡± for the smaller) for noise and this is for small high frequency ¡°noise¡±
let me know what you think?
steve
|
Re: Super low noise PSU for the OXCOs
I just remembered that yahoo ate a project explanation that I'd spent 2 hrs writing. I walked away in disgust and forgot
to address the gap. Supply is to feed 4 of the CTI OXCOs.
Basic supply:
3:1 stepdown transformer at 120 nominal Vrms in 40 Vrms out CT ~60 VA 27 Vp
dual rectifier bridge feeding large Farad level capacitor capacitor bank with choke input and EMI filtering.
dual LM317 based regulator at 16 V out (Fig 15)
80-100 F 16 WVDC capacitor bank for load stability
15 V out for the LM399 and the AD8429 instrumentation amps
dual LM317 based regulator with 15 V in and 5 V out for the OXCO
similar 100 F capacitor bank to provide load stability at 5 V
Mains tolerances are +/-5 % in the US. A single LM317 doesn't have the line stability required to suppress that below 1 mV which is why the datasheet shows using a dual LM317 topology to improve line stability By selecting tempcos I should be able to make it very small for a pair.
There will be extensive shielding, filtering and thermal measures taken as needed.
Have Fun!
Reg
|
Re: Super low noise PSU for the OXCOs
Hello,
oh wow. I light up a cigar and grab some popcorn. Looks like this debate is about to get interesting.
Tam
With best regards
Tam HANNA
--
Enjoy electronics, 3D printing and cigars? Join more than 21000 followers on my Instagram at
|
Re: Super low noise PSU for the OXCOs
Dave,
Home Run!
"Personally rather than all the messing around testing 100s LM317s, I would have designed a power supply a voltage reference chip used in 6.5 digit multimeters - the LM399 "
--------------------------------------------------------------------------------------------------------------------------------
It deftly demonstrates you have never repaired a bench grade DMM. Nor have you read the LM399 and LM317 datasheets. For that matter, you've not bothered with the topic context either. As it is building a PSU for an LM399.
Where do *you* get stable power for high precision long term (multi-year) measurements? I invite you to enlighten us as to how you address the +/-5 % mains tolerances and their ensuing effects. Rotating convertors are fabulous, but rather expensive to operate.
The LM317s are to regulate the mains power to an LM399, one or more CTI OSC5A3B02 OXCOs and the instrumentation amplifier voltage followers. Among other things. I need to reduce the 1e-1 grid variation and *all other* external unpredictable perturbations to well under 1e-9 to 1e-12. Per year!
That requires at least 2 LM317s (e.g. following the preregulator arrangement in the datasheet). I'm likely to do 3 or 4. Dual LM317 circuit from 27 Vp transformer nominal to 15 Vdc with large (100 Farad) capacitor bank to provide load stability followed by 10 V output from 3rd LM317 for the LM399 and the voltage followers with a similar 10 WVDC capacitor bank. Similar feed and topology for the 5 V supply for the OXCO and ATMega. Very heavy filtering and isolation throughout.
By careful selection from data measured over a range of voltages and temperatures, I should be able to make up pairs with as close to a zero tempco as can be done. I'll let an MCU deal with the last few bits.
The tempco matching may be an argument for dual LM317s for each output voltage. I spent much of the day contemplating using a Peltier junction to create a pair of reversible tempco test tunnels.
Have Fun!
Reg
On Wednesday, September 6, 2023 at 04:41:39 PM CDT, David Kirkby <drkirkby@...> wrote:
On Tue, 5 Sep 2023 at 23:22, Reginald Beardsley via <pulaskite= [email protected]> wrote: The LM317s arrived.? Now I? have to characterize them.? The tabs are the outputs so I've ordered nylon threaded hex spacers? to assist in mounting them.
I've got 5x 44421/2A relay cards and a 3457A.? Inclination is to use an ATMega to drive the relay cards via GPIB along with the 3457A.? At NPLC 10 the 3457A will? read each device twice a minute at 7.5 digits in MATH HI-RES mode.? 15 mV seems a good choice of reference voltage as that will put the 3457A mid range on its lowest scale.
I dont know exactly what you are doing , so maybe the 30 mV range is best,?but the worst range for 24 hour stability on the 3457A is its 30 mV range. It¡¯s more than 10x worse than the 3 V range and twice as bad as the 300 V range.?
The relevant question is what are the "ideal" parameters?? I have no clue at this point other than simple stability over time and temperature.? That may be it.? Pick pairs which have best lab environment stability.
There¡¯s a good video on YouTube
of someone evaluating the performance of the ADR1000A?
which is a voltage reference device which?maybe better than the LTZ1000A
?
I say maybe better, since he will not know until years of testing would have been completed.?
Personally rather than all the messing around testing 100s LM317s, I would have designed a power supply a voltage reference chip used in 6.5 digit multimeters - the LM399.?
Have Fun!
Reg
Dave
|
Re: Super low noise PSU for the OXCOs
On Tue, 5 Sep 2023 at 23:22, Reginald Beardsley via <pulaskite= [email protected]> wrote: The LM317s arrived.? Now I? have to characterize them.? The tabs are the outputs so I've ordered nylon threaded hex spacers? to assist in mounting them.
I've got 5x 44421/2A relay cards and a 3457A.? Inclination is to use an ATMega to drive the relay cards via GPIB along with the 3457A.? At NPLC 10 the 3457A will? read each device twice a minute at 7.5 digits in MATH HI-RES mode.? 15 mV seems a good choice of reference voltage as that will put the 3457A mid range on its lowest scale.
I dont know exactly what you are doing , so maybe the 30 mV range is best,?but the worst range for 24 hour stability on the 3457A is its 30 mV range. It¡¯s more than 10x worse than the 3 V range and twice as bad as the 300 V range.?
The relevant question is what are the "ideal" parameters?? I have no clue at this point other than simple stability over time and temperature.? That may be it.? Pick pairs which have best lab environment stability.
There¡¯s a good video on YouTube
of someone evaluating the performance of the ADR1000A?
which is a voltage reference device which?maybe better than the LTZ1000A
?
I say maybe better, since he will not know until years of testing would have been completed.?
Personally rather than all the messing around testing 100s LM317s, I would have designed a power supply a voltage reference chip used in 6.5 digit multimeters - the LM399.?
Have Fun!
Reg
Dave
|
Re: Super low noise PSU for the OXCOs
The LM317s arrived. Now I have to characterize them. The tabs are the outputs so I've ordered nylon threaded hex spacers to assist in mounting them.
I have a nice 20 V 5 A linear lab supply to feed the parts and a precision voltage reference and op amp to drive the Vctl pins and the metering system. And a suitable means of controlling the environment.
I've got 5x 44421/2A relay cards and a 3457A. Inclination is to use an ATMega to drive the relay cards via GPIB along with the 3457A. At NPLC 10 the 3457A will read each device twice a minute at 7.5 digits in MATH HI-RES mode. 15 mV seems a good choice of reference voltage as that will put the 3457A mid range on its lowest scale.
I've been wrestling with the how to measure thermal stability. A circulating system with a set of Peltier devices and lots of thermal sensors and muffin fans seems a start.
I just cut a pair of 83 mm squares of 0.200" aluminum which I'll glue with thermal cement to opposite sides of four 40 mm Peltier junctions which will in turn be attached to large heatsinks. The plates are because the heatsinks are only 70 mm wide. It should improve transfer from the Peltier device to the heatsinks. That will let me measure them heating and cooling just by reversing the Peltier polarity. A hot loop and a cold loop. Muffin fan apiece. There's no need for strength, so roof flashing will serve for ductwork. Foil face foam insulation mounted at the corners inside a large box with batting to suppress air currents.
I built a triple stacked junction device with a CPU fan. It will hit freezing in seconds. That's part of a more elaborate version of this environmental test setup.
The relevant question is what are the "ideal" parameters? I have no clue at this point other than simple stability over time and temperature. That may be it. Pick pairs which have best lab environment stability.
Have Fun!
Reg
|
Re: Super low noise PSU for the OXCOs
As there seems to be significant misunderstanding of the problem I am addressing in this thread I shall elucidate:
1) Precise voltage doesn't matter. The tuning voltage will be varied constantly as needed.
2) Low noise and long term stability are the major requirements.
3) Any voltage to any of the devices from the mains to the OXCO supply within device tolerance is perfectly acceptable so long as it is stable. My focus is frequency, not voltage.
4) Aging effects are tractable, but not easy at 1%/1000 hrs. We shall see what I get from ebay.
5) US mains tolerance is +/- 5%, 114-126 Vrms and nominal 118 Vrms. I'm not in the mood to reprise the error propagation from the mains to the OXCO supply. You can read the LM317, LM399 and CTI OSC5A2B02 datasheets and labor thorough that yourself. I am confident that you'll arrive at similar conclusions. If I have made an error please show my error.
What matters is vanishingly low noise and predictable aging behavior from the PS to the OXCOs. Because of device aging, the Vctl voltages will require constant adjustment to achieve 0.001 ppb/yr frequency stability.
This is NOT a precision voltage reference. It is an ultra low noise, high stability power supply. Not at all the same despite shared difficulties and similarities.
Have Fun!
Reg
|
Re: Super low noise PSU for the OXCOs
Mr. Kirkby,
I sent you a private email and deleted your off topic posts from the archive. Asking why I didn't choose a different part with a bunch of smileys is off topic. You have brought the matter up again with simply more arm waving. You are aware of some aspects of the continuum mechanics of a voltage reference. I can write and solve the equations.
In the continuum mechanics of materials, aging falls under the rubric of viscothermoelastoplastic behavior. And that's if you leave out diffusion and other factors. The constituent equations are rather complex to put it mildly. At such time as I have some measured data I shall address them in this application context. I did this professionally in the oil industry dealing with rocks with various fluids in the pore spaces and developed test cases which identified significant errors in commercial software that was quite expensive per seat and came with 5 figure annual maintenance fees.
I should like to apply my experience and mathematical skills to the problem, but neither the time-nuts nor the volt-nuts will provide data except as graphs which are of no use because of the noise in the data. Noise I know how to remove. So I have no choice but to collect my own data.
What's the voltage reference chip in an HP 3457A? Please excuse my not bothering to take mine apart, but I'd bet it's an LM399. I know that's what my 34401As use and probably the 3478A as well.
What would an LTZ1000A offer over an LM399 that would justify the cost? I am not building a precision voltage reference, I'm building a low noise PS for a set of 4 OXCOs which have a power supply frequency variation of 8 ppb/mV according to the datasheet.
I need to feed stable power to the LM399. Where's that going to come from? Battery? That isn't going to work long term because it will discharge and the LM399 output voltage will change. Put a charger on it and it's just a capacitor.
I am describing the design considerations for a project for which I have chosen most of the parts. the main exception is the op amp required for the voltage follower to supply power to the OXCO. The control voltage supply requires minimal current. The power for a *single* OXCO is over 2 Watts. I'm not aware of a low noise op amp that can supply that much power.
I've done an error budget starting at the +/- 5% mains tolerance, calculated the bleed through of a single LM317 for that variation. It is far too large to use only a single LM317. I have a small Sola constant voltage ferroresonant transformer which should allow using a single LM317. But those are expensive and a pair of LM317s appear to be capable of meeting the noise rejection with proper design choices, so I'm testing that also.
My goal is 0.001 ppb/yr. The LM317 1% of Vout per 1000 hours translates to 50 mV per 1000 hours feeding the OXCO 5 V pin. That is 400 ppb/1000 hrs. That is clearly not going to work. My *hope* is that used LM317s will exhibit the same aging behavior as crystals and voltage references. It contains one, so that is not an unreasonable expectation. Thus I expect something more akin to 0.01%/yr aging which will be far easier to address.
Reg
On Sunday, August 27, 2023 at 03:14:14 AM CDT, David Kirkby <drkirkby@...> wrote:
On Sat, 26 Aug 2023 at 21:00, Reginald Beardsley < pulaskite@...> wrote: This is completely off topic. Please refrain from such postings.
It¡¯s not off-topic.
There are very sound technical reasons for selecting used OXCOs and voltage references instead of new parts quite independent of price. They have vastly better long term stability as a consequence of many years of burn-in that would make them completely unaffordable from an OEM as new parts.
In order to get best long term stability from an LTZ1000A voltage reference you need the PCB to be designed to limit stress on the wires, so the board has cutouts. It should be blindingly obvious that desoldering an LTZ1000A is going to reset the aging process.?
If you look on eBay at the used LTZ1000A¡¯s they are often dented. But even if desoldered carefully, you can not expect the stability it had before being desoldered. So there¡¯s no advantage in buying a used LTZ1000A chip than a new one.?
Using a complete used voltage reference board, such as those that show on eBay from 3458As makes more sense for ultimate stability, but they are a few hundred dollars, whereas a new chip is $50.?
If you want a fairly inexpensive voltage reference, the chip used in the 6.5 digit 3457A is inexpensive
The output voltage of an LM317 has a 1% per 1000 hours drift. But no one is going to salvage LM317s so that will probably be my single largest source of long term instability.
If that¡¯s your largest single source of long-term stability, then why use such a chip if you are looking for the ultimate performance as the title implies.?
There are plenty of better chips which will give orders of magnitude lower drift and less noise. I forget what the chip used in the HP 3457A (6.5 digit) meter is, but it not that expensive and orders of magnitude more stable. It has a heater but it doesn¡¯t use much power. The 3457A multimeter doesn¡¯t have a fan.?
Obviously the output of the voltage reference IC would need buffering, but done properly, it should be possible to achieve orders of magnitude lower drift.?
Reg
Dave.?
|
Re: Super low noise PSU for the OXCOs
If you need a PCB dude, chant a devils song to my personal email, not the list.
I can fire up by IBF and whack something together.
With best regards
Tam HANNA
--
Enjoy electronics, 3D printing and cigars? Join more than 21000 followers on my Instagram at
|
Re: Super low noise PSU for the OXCOs
On Sat, 26 Aug 2023 at 21:00, Reginald Beardsley < pulaskite@...> wrote: This is completely off topic. Please refrain from such postings.
It¡¯s not off-topic.
There are very sound technical reasons for selecting used OXCOs and voltage references instead of new parts quite independent of price. They have vastly better long term stability as a consequence of many years of burn-in that would make them completely unaffordable from an OEM as new parts.
In order to get best long term stability from an LTZ1000A voltage reference you need the PCB to be designed to limit stress on the wires, so the board has cutouts. It should be blindingly obvious that desoldering an LTZ1000A is going to reset the aging process.?
If you look on eBay at the used LTZ1000A¡¯s they are often dented. But even if desoldered carefully, you can not expect the stability it had before being desoldered. So there¡¯s no advantage in buying a used LTZ1000A chip than a new one.?
Using a complete used voltage reference board, such as those that show on eBay from 3458As makes more sense for ultimate stability, but they are a few hundred dollars, whereas a new chip is $50.?
If you want a fairly inexpensive voltage reference, the chip used in the 6.5 digit 3457A is inexpensive
The output voltage of an LM317 has a 1% per 1000 hours drift. But no one is going to salvage LM317s so that will probably be my single largest source of long term instability.
If that¡¯s your largest single source of long-term stability, then why use such a chip if you are looking for the ultimate performance as the title implies.?
There are plenty of better chips which will give orders of magnitude lower drift and less noise. I forget what the chip used in the HP 3457A (6.5 digit) meter is, but it not that expensive and orders of magnitude more stable. It has a heater but it doesn¡¯t use much power. The 3457A multimeter doesn¡¯t have a fan.?
Obviously the output of the voltage reference IC would need buffering, but done properly, it should be possible to achieve orders of magnitude lower drift.?
Reg
Dave.?
|
Re: Super low noise PSU for the OXCOs
A minor gloat. Color me amazed. I didn't think anyone would take the trouble to salvage 1.5 A LM317s. But ebay came to the rescue.
I just ordered 110x *used* LM317s in the TO-220 package from CA via ebay for $24 and change delivered.
There is no spec on long term aging, just the datasheet 1% per 1000 hours, so I'll need to test a bunch to pick some with minimal residual drift. The leads were snipped where they thin for the PCB hole, but that's of no consequence as I am also facing a complex mounting job to minimize LM317 temperature variation as the OXCO heaters turn on and off. I'm hoping lots of aluminum will do and that a Peltier device and PID will not be needed for temperature control, but if it is I'll do that.
I'd really like to use a voltage reference rather than the standard arrangement using Vout for the 2nd LM317. Not clear to me at the moment how to do that and meet the adjust line minimum current drain. If anyone has suggestions please post them.
I am acutely aware this is a full blown lunatic project, but I'd like to note that sub uV noise levels are significantly less difficult than 1 ppm voltage accuracy. That is seriously hard as is well documented on EEVblog and many other places.
Have Fun!
Reg
|
Re: Super low noise PSU for the OXCOs
The op amp I selected for the voltage follower for the LM399 is an AD8429BRZ.
It will be challenging for an old guy with poor eyesight and no PCB CAD skills but I should have or be able to get a breakout board for it.
This was the lowest noise part I could find. Very much appreciate comments and advice on the choice.
Reg
|
Super low noise PSU for the OXCOs
I gutted an ancient CB radio PSU yesterday which served as my main 12 V PSU along with a 5 V supply I built from scratch many years when I was in school.? It's a 40 Vrms CT with 22 AWG leads.? It used a bare LM317, so not a beefy transformer.? I considered it 1 A max when I repaired it 50 years ago.
I had been planning to use the existing chassis, but as I contemplate what I want to put in there I think a bigger chassis will save a lot of misery.? So I think the transformer and fuse holder may be the only parts I use.
Stage one will be a brute force full wave rectifier PS putting out about 28 Vdc with an IEC EMI filtered socket for the mains feed and large caps and chokes on the output.? The LM317 datasheet shows -70 dB ripple rejection with a 15 V drop across the LM317.? The LM399 needs 10 Vin so that makes the drop across the 2nd LM317 a bit marginal so I'll have to trim the pair of LM317s to see what produces the least ripple.
For the 10 V feed to the LM399 I plan to use a 5-10 F super cap and similar on the 6.95 V output and also for the PS for the ultra low noise op amp that will actually supply the OXCO Vctl divider.? Not yet sure about the 5 V supply to the OXCO.? At turn on they pull almost 1/2 A which is a lot to ask of an op amp.? Once they are warm they drop to about 1/4 A.
As there has been quite a bit of FUD about the CTIs "jumping" in frequency, I've attached a screen dump from a recent multi-day experiment using a tinyPFA, Bodnar GPSDO, new PCB OXCO,? HP 5386A, Instek MSO2204EA and Instek GPE4323 PSU.? The GPSDO and OXCO were feeding JFW 1:5 splitters so that the DSO, 5386A and tinyPFA all had the GPSDO for the reference signal.? Over the course of 3 days I observed a constant 0.01 Hz (last digit) bobble on the 5386A and a 0.005 Hz bobble on the tinyPFA which was consistent with drift of the signals on the DSO.
I attached the DSO to the PSU using a BNC to dual banana plug cable to capture the screen dump. I don't yet know what the source of the noise burst is. It is periodic and I know that the DSO has major SMPS noise problems and will be replaced with a low time Tek 2465B by the end of the day. The Instek PSU is a linear supply so it should be much quieter than the SMPS in the DSO.
The CTI OSC5A2B02 datasheet specifies +/- 2 ppb for +/- 5% supply variation. That's 8 ppb/mV which translates to over 90 ppb peak shift during the 200 ns noise burst. The voltage regulation on the new PCB OXCO reduces that substantially, but measuring actual supply variation to the OXCO module at that voltage level without external EMI pickup will not be easy and will definitely require significant effort and a custom circuit.
So I plod along in the midst of property management, trying to organize far too much stuff I have bought and sell the mountain of tube era electronics and other stuff Dad left me.
Anyone interested in an NOS Montgomery Ward 3.5 cu ft cement mixer still on the original shipping pallet? FOB Heber Springs, AR.
Have Fun!
Reg
|
Re: CTI OSC5A2B02 OXCO module high precision frequency reference project
I received a tinyPFA from R&L and am struggling to figure out how to use it. The documentation has a long way to go.
I currently have a Bodnar 2 output GPSDO feeding the A input of my 5386A running on its internal OXCO and feeding the A port on the tinyPFA. The difference frequency shown on the tinyPFA is bouncing around at a few mHz, roughly +/- 0.005 ppb which is a long way from my goal. The 5386A shows the same jitter in the last digit.
I stuck my DSO on the power supplying both the GPSDO and the CTI OXCO and took the attached screen dumps. One running and one singleshot.
The lab PSU is an Instek GPE 4323 linear supply. For general use it is excellent and I've been quite happy with it. However, the noise on the supply shows the difficulty I am facing. It is much too dirty to feed the OXCO in either oscillator even though both have internal voltage regulation. I am getting a roughly factor of 200 noise reduction, but I need to make that noise unmeasurable at 1 mV/div. I suspect that the 15 MHz ripple is actually noise pickup from the absolute crap SMPS in the Instek MSO2204EA. I have a Tek 2465B in transit which I shall be using as my primary bench scope in the future until I have the time and inclination to shield the Instek and other DSO SMPSes.
For the benefit of those who have not examined the PS problem I should like to suggest considering the following:
USA 120 mains are +/- 5% (114-126 Vrms, nominal 118 Vrms). Then consider the +/- 5% variation of an LM317 to supply variation. Followed by a 2nd LM317 feeding an LM399 supplying both the 5V CTI power and the Vctl pin divider. For simplicity assume that both LM317s are using an AD584JH voltage reference instead of the output of the LM317 for their reference voltages with a several Farad cap to stabilize the AD584JH feed to the voltage dividers.
Once I have had time to complete a <1 uVrms mean noise level PSU I'll write up a detailed error analysis with full documentation.
Have Fun!
Reg
|
Re: [Question] CTI OSC5A2B02 OXCO module high precision frequency reference project
Grif,
Thanks for your very interesting and insightful question. I've never thought about that, but it suggests an excellent experiment.
The effective capacitance at a particular discharge rate should be measurable using a DSO in single shot to measure the voltage across the battery when a load is applied and then solving for C.
I don't know enough about battery behavior to offer anything more than a SWAG that the effective capacitance varies with the applied load.
This might be an excellent case for a battery of Edison cells for a low noise PSU.
Have Fun!
Reg
On Friday, August 11, 2023 at 03:00:12 PM CDT, Grif Griffith <grif615@...> wrote:
Not so much design as a hobby, but my noise floor really shrunk when I powered my station with a deep cycle battery. Wonder what the equivalent capacitance is on a group 24??? ;-) the switching smart charger is not noticeable with the battery in line.
|
Re: [Question] CTI OSC5A2B02 OXCO module high precision frequency reference project
Not so much design as a hobby, but my noise floor really shrunk when I powered my station with a deep cycle battery.? Wonder what the equivalent capacitance is on a group 24??? ;-)? the switching smart charger is not noticeable with the battery in line.
On Aug 11, 2023 10:50, "Reginald Beardsley via groups.io" <pulaskite@...> wrote:
Jerry,
Thanks as always. That's very helpful, though ti adds a 3rd option to test.. This seems like a good application for 1-2 Farad caps on all the DC lines adjacent to the device being powered.
That should yield sub 1 Hz RC time constants. As Ken Thompson said, "When it doubt, use brute force."
One of the things that scares me is EMI. A "make before break" relay to switch between mains supply and a battery for measurements.
At the moment I'm swamped by the consequences of a buying binge arriving.
Have Fun!
Reg
On Friday, August 11, 2023 at 11:59:02 AM CDT, Jerry <jerry@...> wrote:
This person named Adrian Rus, pretty sharp engineer and entrepreneur, (co-authored some papers with Ulrich) said (and we tested it) that you should have a minimum of 15,000uf on an individual oscillator for references. Also, and I tested this for him, we switched to lt3045 regulators as they were proven to be exceptionally low-noise and cleaned-up even dirty power supplies. We used these modules and others like them:
https://www.ebay.com/itm/274137165726?mkcid=16&mkevt=1&mkrid=711-127632-2357-0&ssspo=pX50eYVtSOq&sssrc=2047675&ssuid=UQSWHCvkTY2&widget_ver=artemis&media=COPY
For testing, I used a 60dB amp that tapped off the residual noise and A/C from the DC voltage and then fed that into my low frequency spectrum analyzer and the noise reduction was amazing.
Andrew Holme, who authored our phase noise test set, used a capacitance multiplier circuit with large caps on it. I tested it and then fed it into the lt3045s and it was pretty clean. You can stack the lt3045s for more current and also in series to reduce noise.
Big problem turned out to be shielding when testing very low PN oscillators in the -180 dBc range so for testing I switched to Dewalt 20V power tool batteries followed by the lt3045 regulators. When using batteries and 15kuf caps though, I first used the LM317 regulators and didn¡¯t see much difference but running off of mains, the lt3045s made all the difference. Every noise spike we found later was related to external energy.
If I was building a bench, I think I would have a rack of the lt3045s with a switch that fed them from batteries or from mains.
Sent from Mail for Windows
|
Re: [Question] CTI OSC5A2B02 OXCO module high precision frequency reference project
FYI I just ordered 4 from this seller:
https://www.ebay.com/itm/115745541000
Which was about 1/4 of the other seller's prices. Makes them suspect, but Shenzen is truly the Wild West and Far East.
I'll test and post results when they arrive.
Have Fun!
Reg
BTW Trimming everything after the "?" deletes ebay's tracking information. I alsways test the link after the edit to verify I didn't drop a digit.
-------------------------------------------------------------------------------------------------------------------------------------
On Friday, August 11, 2023 at 11:59:02 AM CDT, Jerry <jerry@...> wrote:
This person named Adrian Rus, pretty sharp engineer and entrepreneur, (co-authored some papers with Ulrich) said (and we tested it) that you should have a minimum of 15,000uf on an individual oscillator for references. Also, and I tested this for him, we switched to lt3045 regulators as they were proven to be exceptionally low-noise and cleaned-up even dirty power supplies. We used these modules and others like them:
https://www.ebay.com/itm/274137165726?mkcid=16&mkevt=1&mkrid=711-127632-2357-0&ssspo=pX50eYVtSOq&sssrc=2047675&ssuid=UQSWHCvkTY2&widget_ver=artemis&media=COPY
[snip]
|
