Lars DIY GPSDO with Arduino and 1ns resolution TIC

[MIS42N]

I replaced it with one of these eBay units, item 156270499091, a generic adjustable buck converter using an LM2596.

That's the same design buck converter as I settled on. Set to 7V followed by 100uH and 470µF LC filter then 2 * LM1085 low dropout linear regulators. One for the OCXO, one for the processor and ancillaries.

[Johnny B Good]

Johnny I use a common LM7805 with heatsink in my Lars/Murray GPSDO. The ovened oscillator is 12 Volt with pretty good regulation.

Regarding buck converters and potential noise, I had the 1.2 Volt linear regulator in my QS1R SDR receiver fail (it supplies a field programmable gate array)... a not uncommon failure as the designer didn't check the current/power dissipation when running 7 "slices", or separate bands, simultaneously, a replacement got hot enough to make toast. I replaced it with one of these eBay units, item 156270499091, a generic adjustable buck converter using an LM2596. Given the nature and use of the SDR receiver, covering 9 KHz to 60 MHz, I was surprised to hear no "hash" from the board. In the past I have tried an encapsulated 12V to 5V unit intended to be used for USB outlets in vehicles and found it plain unusable due to high levels of noise all the way from HF to UHF... It carried no CE or "e" mark for fitting permanently in vehicles.

SJ

 Thanks for that information,

 When I finally decided to transfer my 13MHz ocxo based breadboarded prototype onto a PCB (a 100mm square stripboard that would neatly fit and slide into a 50mm tall extruded aluminium project box), I did toy with the idea of using a 7805 to supply all the logic, the M8N and my ocxo from a single 5.3v rail but the need to bolt it onto the obvious heatsink (aka, the aluminium case) didn't appeal to me.

 Not having an actual datasheet (or variable bench supply at that time), I tested the ocxo on a 5v supply (actually 4.8v from the Y robot breadboard power adapter) rather than assume it was a 12v version and risk blowing up my one and only 'precious' ocxo should it prove not to be a 12 volt type.

 Starting it up from 4.8v required a few seconds for the 13MHz square to appear out of the noise (using a 5v supply had it spring into life within less than a second) but it did provide a stable 4v pk-pk signal that I could use to feed my 1.3 divider logic ( a doubler feeding a divide by 13 TTL stage which then fed a second  5x clock multiplier with its specified minimum of 2Mhz to generate a 10MHz locked to the ocxo's 13MHz.

 As you may imagine, a bunch of TTL beavering away to generate 10 MHz from a 13MHz clock source built onto strip board is a rather noisy combination. When using the FFT spectrum analyser feature of my SDS1202X-E to monitor the Vcc rail now powered from a 1.3A mini 360 buck regulator to see just how big a problem its claimed 30mV pk-pk ripple could be was a pleasant surprise.

 Well, I have to say, I was only able to find its contribution to all the TTL hash on the power rail by knowing that it switched at 1.2MHz. It was barely visible amongst all that noise which didn't seem to trouble the operation of my MK I gpsdo (and which would have been present regardless if I had used an analogue regulator instead). From that moment on, I knew I would never be using such analogue regulators in any of my future gpsdo projects.

 This experience with that (as I was later able to prove) 12v 13MHz ocxo is what led me to powering the 10MHz ocxo in the MK III with just 10.25v having proved that this was sufficiently above the minimum (a surprisingly high 8 to 9 volts threshold) required to satisfy the internal reference voltage regulator which powered the oscillator and the micro-controller in charge of the oven heater. I did this to reduce the initial cold start warm up demand which would also have to supply an additional initial 300mA charging current to the backup LiPo cell in the event of a 'black start' worst case scenario.

 Regarding the noise produced by DC-DC converters, I whole heartedly agree with your experience, In my initial attempts to source such converters with an acceptably low level of ripple and switching noise, these 7805 sized mini 360 buck regulators have been the exception to my own dire experience with the larger examples. In one case, I  saw 800mV of ripple which seemed to be due to a missing rectifier diode which only dropped to 400mV of ripple when I soldered a diode into the vacant position.

 In desperation, I even tried out one of those "Two in One" buck converters which included a 5 volt ldo regulator in what I rightly guessed would be a doomed attempt to get rid of the buck converter's switching ripple and noise. I have to say I wasn't disappointed to have my guess confirmed but it would have been quite nice to have been spared such disappointment in this case. If anything, the noise was almost as bad as that of the missing diode specimen!

 Apropos of this being, that following the output of a buck converter outputting 8 volt into a typical 5 volt analogue voltage regulator will offer little to no benefit in reducing switching ripple and noise since their 80dB ripple rejection is only achieved with 100 and 120 Hz supply ripple frequencies tailing off to almost nothing at 10KHz, above which they become transparent to such noise. There are speciality analogue regulators available that can provide ripple rejection ratios of 20 to 30dB at 100KHz but I believe these are intended to clean up the output of mains smpsus soft switching at 30 to 60KHz.

 Looking at that moduleme example you mentioned, imo, I think  the Banggood mini 360 7805 sized 2.1 A module would be an even better alternative solution, assuming you're not feeding it from a supply rail above its 24v maximum rated limit and the load is within its 2.1A maximum continuous limit (3A peak).

[Johnny B Good]

 It's been bugging me as to why anyone would prefer the bulkier/noisier buck converter modules over the rather elegant/excellent mini360 modules that, seemingly, only Banggood (and possibly AliExpress - I haven't checked) sell and I think what's putting people off, is the use of a rather cheap and tiny variable resistor that makes accurate and stable voltage adjustment almost impossible to achieve.

 I guess I should have mentioned that I usually unsolder this 'token of adjustability' and use the solder blob links instead, perhaps, as I've done when I want to set the 5v option an extra 250 to 300mV higher by bridging the 5v solder blob jumper with a low value (circa 1.8K) resistor. The only exception this time round being when I wanted an output voltage of 3.9v to safely float charge a protected Lipo pouch cell where I bridged to 5v link and used the variable resistor to fine tune the output voltage.

 I'm in two minds over whether to leave it as is or else remove it, measure its resistance and fit a matching resistor in its place. On the one hand, it's stable enough as is in this case and allows me to further fine tune the voltage and, should the worst happen (like with one those larger modules using an even shoddier variable resistor) and it falls apart and go open circuit, the only casualty will the protected LiPo if the fault goes unnoticed for more than a few weeks. The voltage will only rise to the jumpered voltage which is safely within the input voltage range of the boost converter which supplies 10.25v to the ocxo and input to the second buck converter generating the 5,3v Vcc for the logic and the ZED9 module.

 The only other time, quite recently, where I've relied solely upon this variable resistor to set the output voltage, was when I was testing whether I could power an LPRO 101 with a stable and regulated 21v from a 24v psu. In this case I used my SDL1020 rather than the LPRO itself. I could only get it to max out at 20.9v (close enough in this case) and mounted it to a heatsink to improve its reliability at the initial cold start current of circa 1.7A which settles down to under half an amp once up to temperature some 4 or 5 minutes later.

 This test was rather encouraging (in that it didn't blow up after running it at 2A out with a 26v input for twenty minutes or so). I haven't, as yet taken this beyond the testing stage. It's just an idea with the aim of minimising heat loading in the final build of my rubidium lab frequency reference to improve stability at the upper 32 deg ambient temperature limit of the +/- 1 to 5mK stability of base plate thermal regulation set to 36.050 degrees. The low end of the ambient temperature range appears to be somewhere below the -2 degree mark where the base plate remains within 1mK of the set temperature from around 18 degrees downwards.

 I prefer these tiny and efficient buck regulators on account of their superior voltage stability with temperature over all the alternatives I've tried (boost and buck boost converters) to stick to my use of the rather plentiful supply of commodity 19v laptop charging bricks as my go to external dc power source. In the end, I decided to go for an impressively low noise 24v 2A continuous / 3A peak mains psu board and build my own "Power Bricks", hence my interest in using those cute little mini 360 buck converters in my rubidium frequency standard.