On 9/18/21 10:50 AM, Lux, Jim wrote: > LT3042 and LT3045 are your friend here.  Excellent PSRR well up past 10 > MHz. > > If you need more current, you can just parallel them. > > $6 each for the 3045 from Digikey For best performance, take a look at the layout suggestions in the 3042/3045 datasheet, and especially the way the evaluation board is laid out. There's quite a difference between their recommendation and what my first uneducated attempt looked like. John

Did you say there's a 1 mH choke at the LDO input? I gather that overall, you're regulating 5 V in to 3.3 V out, running the IC that takes a few mA DC, and loading the CMOS output with 50 ohms scope input (or a 50 ohm terminator on a scope or probe setup). I'd recommend the following changes. 1. Get rid of the choke, or at least replace it with a much smaller one that will be more effective - say no more than 100 uH. Especially in surface mount, it will be much smaller too. 2. Put big bypass caps at the input and output of the LDO. Solid Ta caps are best for this, like in the tens to a hundred uF, readily available in surface mount. If you keep a choke at the input, bypass the 5 V input end also. Especially for experimenting, it's a good idea to add reverse protection in case you screw up the hookup. For low voltage stuff, use a shunt diode from the input supply to ground, big enough to dump the supply if backwards. 3. Don't bother with DMMs and scope-probing the various supplies and such, except for troubleshooting. Once you confirm proper operation, disconnect all the probes and grounds - otherwise, they'll just form all sorts unnecessary ground loops, and confuse things. They don't have good enough CMRR to show what's really going anyway. 4. Run the I/O signals in coax. Ultimately, the actual application may need the full output signal, but for experimenting, it's best to pad the output into coax to get it to the scope or SA, or whatever. A simple method is to AC-couple it with a good RF cap, say in the 10 nF range (for 26 MHz), then a series resistor into the coax, and a shunt resistor there, to get it in the 50 ohm Rs range - it doesn't have to be perfectly matched. The series resistor should be as big as possible, to minimize loading on the IC output, but small enough to get sufficient signal for measurements. Again, in an actual possible application, the full signal may be needed, but the load on the output would only be the input of a buffer amp or logic, that's nearby, so not needing to take a long trip. This setup should provide a more realistic view of how the IC behaves. For the control voltage, there would not be RF issues, but some bypassing at the IC is good, just to clean up local RF, but not so much that it affects loop response if you're PLL-ing it, for instance. There can be DC ground loop issues though. Looking at it as a black box, you'll have the power supply common, the output common, and the control signal common, all capable of carrying some of the supply return current, depending on how things get hooked up. In a real application, these paths would be minimized and dealt with, but for experimenting, you have to keep it in mind and be sure you know and properly control the tuning voltage, referenced to the local common at the IC. It's also a good idea (for experimenting) to clamp the tune voltage within its normal range, just in case. This is not likely needed in a real application. This should provide a fairly clean setup. If there are still annoying common-mode issues with the measurements, you can add CM chokes as needed, around all of the lines together, going to the black box. Ed

On 18/9/21 11:12 pm, Julien Goodwin wrote: > > > On 18/9/21 10:26 pm, Bob kb8tq wrote: >> Hi >> >> The chip is intended to be used with the divider engaged ( set to a divide of >> 2 or greater). That should act as a pretty good buffer if the layout is reasonable. > > I'm using the -17 variant, no divider, running at the 26MHz fundamental. > >> Looking at the spectrum analyzer plots, you either have crud on the control >> line ( ground it and see what happens … ) or on the supply. Either way it’s at >> audio frequencies. Your regulator may have issues (switchers are not what you >> feed a crystal oscillator with …..) or you need some caps in the 100’s of uf >> range on the regulator you have. > > First stage input is a Keysight N675xA supply, not the absolute quietest > supply out there but plenty good enough (it's also currently the best I > have, my last two linear supplies died, and I've not replaced them), the > second stage is an MCP1802 LDO, now with 1mH on the input in series. I'd > have expected 10uF to be enough bulk capacitance, but I can absolutely > try chucking some more on and seeing if that does help. > > When I was just running the DAC without the oscillator the lines were > dead quiet, I suppose it's also possible there's leakage from the > control signal running too close to one of the lines to the crystal > (just a hair under 1mm between the traces) > >> Pay attention to the max output C even when running the divider. You can drive a >> scope probe, but not a 50 ohm line with the device. For 50 ohms you will need >> a pretty healthy ( = high current ) buffer. > > It /should/ be able to handle driving into 50 ohms per the spec sheet, > but I agree it's not doing a great job of it. ... no it shouldn't, my ability to do mental ohms-law is apparently terrible, thanks for those correcting me off-list. Some more experimenting today actually pointed out that the output clears up almost entirely once I get the breadboard out of the picture, which I guess isn't too surprising, but is annoying. I don't normally use breadboards at all, as I'm normally working with SMD parts and it's often quick to knock out a partial board as a test. I'll design up a carrier PCB with proper connectors for power & output, I'll also chuck a buffer on for driving 50-ohm targets as a second output which will help make clear if it's worth integrating one for general use. >>> On Sep 18, 2021, at 12:38 AM, Julien Goodwin <time-nuts@studio442.com.au> wrote: >>> >>> On 13/9/21 6:31 pm, Julien Goodwin wrote: >>>> https://www.microchip.com/en-us/product/PL500-16 (there's various other >>>> versions depending on the frequency you're after) >>>> >>>> Haven't seen any discussion about this on-list, but the PL500 is an >>>> easily (well, in normal times) available VCXO control chip, for those >>>> who might want to make their own disciplined oscillator, especially at >>>> less standard frequencies. I had some arrive today and put the board >>>> I've designed as an OCXO, and was able to trim +/- ~3kHz (around 26MHz >>>> nominal in my case, pretty much the expected +/- 150ppm), all really easily. >>>> >>>> I can't yet say much about quality as it turned out I'd put the wrong >>>> regulator footprint on the board, and with no local regulation the power >>>> rail was jumping all over the place, once I actually fix that and >>>> hopefully get it mounted in its intended enclosure for thermal control >>>> it'll be interesting how it goes (yes this was the project I was hoping >>>> to use my SR620 to monitor the other week). >>> >>> The thermal and shielding situation is to improve, but I did at least >>> get local regulation fixed, and while improved, it's still not great. >>> >>> With a 1mH inductor on the input (pre-regulator) and the local regulator >>> installed: >>> https://twitter.com/LapTop006/status/1439081534053515266 >>> >>> Traces are: >>> Yellow - Output signal (50-ohm terminated) >>> Green - Control voltage >>> Blue - 3.3v rail (main internal rail) >>> Red - 5v input rail >>> >>> At a rough guess I either need more bulk capacitance on the 3.3v rail, >>> or, more likely, lower impedance decoupling caps (I'm currently using >>> 100n 0603 of the "whatever I have in stock" variety). Would welcome >>> suggestions. I /do/ have an impedance analyzer that can handle this >>> frequency (goes to 500MHz), but I lack the SMD text fixture for it. >>> >>> I suspect an output buffer would really help too, and on its own might >>> significantly improve things. >>> _______________________________________________ >>> time-nuts mailing list -- time-nuts@lists.febo.com -- To unsubscribe send an email to time-nuts-leave@lists.febo.com >>> To unsubscribe, go to and follow the instructions there. >> _______________________________________________ >> time-nuts mailing list -- time-nuts@lists.febo.com -- To unsubscribe send an email to time-nuts-leave@lists.febo.com >> To unsubscribe, go to and follow the instructions there. >> > _______________________________________________ > time-nuts mailing list -- time-nuts@lists.febo.com -- To unsubscribe send an email to time-nuts-leave@lists.febo.com > To unsubscribe, go to and follow the instructions there. >

On 20/9/21 5:37 pm, Julien Goodwin wrote: > > > On 18/9/21 11:12 pm, Julien Goodwin wrote: >> >> >> On 18/9/21 10:26 pm, Bob kb8tq wrote: >>> Hi >>> >>> The chip is intended to be used with the divider engaged ( set to a divide of >>> 2 or greater). That should act as a pretty good buffer if the layout is reasonable. >> >> I'm using the -17 variant, no divider, running at the 26MHz fundamental. >> >>> Looking at the spectrum analyzer plots, you either have crud on the control >>> line ( ground it and see what happens … ) or on the supply. Either way it’s at >>> audio frequencies. Your regulator may have issues (switchers are not what you >>> feed a crystal oscillator with …..) or you need some caps in the 100’s of uf >>> range on the regulator you have. >> >> First stage input is a Keysight N675xA supply, not the absolute quietest >> supply out there but plenty good enough (it's also currently the best I >> have, my last two linear supplies died, and I've not replaced them), the >> second stage is an MCP1802 LDO, now with 1mH on the input in series. I'd >> have expected 10uF to be enough bulk capacitance, but I can absolutely >> try chucking some more on and seeing if that does help. >> >> When I was just running the DAC without the oscillator the lines were >> dead quiet, I suppose it's also possible there's leakage from the >> control signal running too close to one of the lines to the crystal >> (just a hair under 1mm between the traces) >> >>> Pay attention to the max output C even when running the divider. You can drive a >>> scope probe, but not a 50 ohm line with the device. For 50 ohms you will need >>> a pretty healthy ( = high current ) buffer. >> >> It /should/ be able to handle driving into 50 ohms per the spec sheet, >> but I agree it's not doing a great job of it. > > ... no it shouldn't, my ability to do mental ohms-law is apparently > terrible, thanks for those correcting me off-list. > > Some more experimenting today actually pointed out that the output > clears up almost entirely once I get the breadboard out of the picture, > which I guess isn't too surprising, but is annoying. I don't normally > use breadboards at all, as I'm normally working with SMD parts and it's > often quick to knock out a partial board as a test. > > I'll design up a carrier PCB with proper connectors for power & output, > I'll also chuck a buffer on for driving 50-ohm targets as a second > output which will help make clear if it's worth integrating one for > general use. Between various shipping delays and my own schedule I only got around to soldering up one of these boards yesterday, only to discover that the sockets I'd been intending to use wouldn't work. A quick trip to the local electronics hobbyist store today to pick up some replacements and I was in business. I'm still not happy with the amount of ripple on the 3.3v supply, with ~330mV peak to peak, however everything else seems solid and the output is much cleaner through a local buffer that can actually drive 50-ohm loads. Sadly, and fittingly for this project where every few steps forward seems to take a step back, just after turning it on today something (probably a cap from the smell) has gone inside the 53310A and so I won't have a clear picture of output stability until that's resolved. For those who want pictures: https://twitter.com/LapTop006/status/1449616018846871558 Then I get to build some more test units and ovenise them. I have the cans and heaters, so this should be easy once I've managed to build some more units with working temperature sensors. >>>> On Sep 18, 2021, at 12:38 AM, Julien Goodwin <time-nuts@studio442.com.au> wrote: >>>> >>>> On 13/9/21 6:31 pm, Julien Goodwin wrote: >>>>> https://www.microchip.com/en-us/product/PL500-16 (there's various other >>>>> versions depending on the frequency you're after) >>>>> >>>>> Haven't seen any discussion about this on-list, but the PL500 is an >>>>> easily (well, in normal times) available VCXO control chip, for those >>>>> who might want to make their own disciplined oscillator, especially at >>>>> less standard frequencies. I had some arrive today and put the board >>>>> I've designed as an OCXO, and was able to trim +/- ~3kHz (around 26MHz >>>>> nominal in my case, pretty much the expected +/- 150ppm), all really easily. >>>>> >>>>> I can't yet say much about quality as it turned out I'd put the wrong >>>>> regulator footprint on the board, and with no local regulation the power >>>>> rail was jumping all over the place, once I actually fix that and >>>>> hopefully get it mounted in its intended enclosure for thermal control >>>>> it'll be interesting how it goes (yes this was the project I was hoping >>>>> to use my SR620 to monitor the other week). >>>> >>>> The thermal and shielding situation is to improve, but I did at least >>>> get local regulation fixed, and while improved, it's still not great. >>>> >>>> With a 1mH inductor on the input (pre-regulator) and the local regulator >>>> installed: >>>> https://twitter.com/LapTop006/status/1439081534053515266 >>>> >>>> Traces are: >>>> Yellow - Output signal (50-ohm terminated) >>>> Green - Control voltage >>>> Blue - 3.3v rail (main internal rail) >>>> Red - 5v input rail >>>> >>>> At a rough guess I either need more bulk capacitance on the 3.3v rail, >>>> or, more likely, lower impedance decoupling caps (I'm currently using >>>> 100n 0603 of the "whatever I have in stock" variety). Would welcome >>>> suggestions. I /do/ have an impedance analyzer that can handle this >>>> frequency (goes to 500MHz), but I lack the SMD text fixture for it. >>>> >>>> I suspect an output buffer would really help too, and on its own might >>>> significantly improve things. >>>> _______________________________________________ >>>> time-nuts mailing list -- time-nuts@lists.febo.com -- To unsubscribe send an email to time-nuts-leave@lists.febo.com >>>> To unsubscribe, go to and follow the instructions there. >>> _______________________________________________ >>> time-nuts mailing list -- time-nuts@lists.febo.com -- To unsubscribe send an email to time-nuts-leave@lists.febo.com >>> To unsubscribe, go to and follow the instructions there. >>> >> _______________________________________________ >> time-nuts mailing list -- time-nuts@lists.febo.com -- To unsubscribe send an email to time-nuts-leave@lists.febo.com >> To unsubscribe, go to and follow the instructions there. >> > _______________________________________________ > time-nuts mailing list -- time-nuts@lists.febo.com -- To unsubscribe send an email to time-nuts-leave@lists.febo.com > To unsubscribe, go to and follow the instructions there. >