As for both yours and DaveR's view that there's no need to fit a cooling fan if either a more efficient smpsu is fitted or else an R type transformer is used for an analogue PSU replacement, I feel I must point out that you're overlooking the fact that both the FY6600 and Fy6800 models were already running uncomfortably hot before any additional heat sources were introduced into the box.
The 25W rated smpsu might be cooler running from its point of view but the fact remains that its efficiency at a relatively low loading may be no better than the crappy original and quite likely adding a little more heat into the box. R type transformers in an analogue PSU aren't the problem, it's the waste heat from the analogue regulators that poses the real threat. R transformers are a good way to go but I'm not so sure about the analogue aspect unless it's being used to eliminate the high frequency ripple of a switching psu with consideration to the penalty of additional waste heat this introduces.
I have to admit being rather puzzled by your statements about "running uncomfortably hot" initially, Johnny, as neither my 6600 nor my 6800 has ever run anywhere near a temperature I'd describe as hot. "Moderately warm" is how I'd describe it, but perhaps it's a subjective assessment? The biggest source of heat, by far, after I did the PS mods in the 6600 was from the 7805 regulator, which got to around 75C and created a local hot spot on the top of the case, but I cured this by swapping it for an OKI 78SR regulator, which runs as cool as the rest of the PS. Any heat inside the case now comes from the Cyclone chip and the 3095 op amps, but even then I need to run them very hard for a couple of hours (something I've only done once, as a test) to get them to an elevated temperature. (In real world usage I rarely use more than about 1Vp-p output, so everything stays very much cooler.) In the same test the D75J TCXO had no problem in coping, with only a 0.1Hz drift at 10MHz after the first 10 minutes of warm up, and no drift after the next 15 minutes - pretty much in line with what you'd expect from an OCXO, depending on the level of sub-Hz accuracy you want.
I still haven't bothered to modify the 6800's PS or op amps even though I've had all the parts to do it since I bought the generator about a year ago - the D75J is the only change I've made, as the awful drift caused by the original XO just had to go. The distortion produced by the original op amps is minimal at the levels I use, so I can live with them, and they run cooler than the the 3095 upgrades so heat produced in the case is even less than that in the 6600. Looking back, the 6600 mods were done more for the challenge than out of necessity, but the 6800 works well enough as it now stands, so I'll probably wait until they do become a necessity before I do anything else to it.
Finally, it's nice to see you back Arthur, and thanks for the update on your mods!
Regards,
Dave
Hi Dave,
Apologies for the delayed response but that was the day me and the XYL started a ten day cruiseship holiday and it's taken me two and a half days to catch up with my TV recording schedule (downloading and converting the iPlayer downloads for future viewing rather than actually watching them!) and sorting out ebay problems with a couple of sellers (including hgfurniture2018's failure to successfully ship that bargain FY6600-60M I'd paid £40.41 for two months ago in spite of a second attempt (allegedly) to ship me another one).
The problem with the FY6600 is that the ventilation slots do very little as far as passive cooling is concerned, even to the extent that tilting it up on its bail makes the situation a little worse. In essence, most of the heat escapes via conduction through the plastic case rather than a more desirable exchange of interior air with that of the cooler ambient air.
This means that the interior air has to absorb the heat dissipated by the components to raise its temperature sufficiently to warm the plastic casing to set up a flow of heat to the outside surfaces in order for the ambient air to conduct the heat away. All these additional thermal gradients between the actual semiconductor junctions generating the heat and the outside thermal environment result in much higher component temperatures than would result from a modest flow of cooling air produced by even a small (50mm square 10mm deep) 12v cooling fan running off the 5v rail.
A simple solution to improving the passive cooling would be to upend the generator to stand it upon its right hand side where the case orientation would provide a chimney effect to speed up the thermo-siphon induced flow of the interior air to improve this transfer of heat to the outside surface of the case (with perhaps some small boost of exchange of air between the interior and exterior of the case). However, the improvement would still be rather marginal for the inconvenience involved, making the addition of a cooling fan the clear winner.
When I remembered that I actually possessed an IR thermometer and used it to check component temperatures, I found that the smd XO ic was running at 50deg C and the three nearby LDO voltage regulators running at 70deg C. The FPGA showed 50deg C as did the base of the heatsink. Since, out of necessity, I had to take these readings with the lid removed, I've no doubt these temperatures would have been a lot higher again in normal use (possibly another 10 to 20 degrees higher but I still don't possess a suitable thermal probe to measure with the lid in place).
I've no doubt that my initial attempt to raise the 12v rails above the 11.7v mark by boosting the 4.97v to 5.49v was adding to the heat stress via the LDO regulators (probably by another 5 deg C) exacerbating this issue a little but even so, a slight 5 degree drop in temperature would still leave the innards on the wrong side of the 50 degree mark with regard to component service life (notably the electrolytic caps which are short enough lived to begin with).
A cooling fan seemed to me an essential necessity, especially when considering that, in the interests of frequency stability, it would likely be spending more time powered up than powered down (good from the point of view of reducing thermal cycling stress induced fatigue but bad for the service life of the caps).
When I do finally get hold of a suitable temperature probe, I won't be in the least surprised to see something like a 15 to 25 degree drop in internal air temperature between fan running and fan stopped test conditions.
In the interim, I have the fact that before fitting the fan, I could detect three notable hot spots, two on the top of the case (PSU and heatsink locations) and one on the underside (heatsink location) the top two of which have now disappeared with the underside one now downgraded to a warmish spot to indicate the effectiveness of my fan cooling upgrade.
I have been rather mindful of the need for quiet efficient effective cooling since before the turn of the century when building desktop PCs. It's all too easy to overlook the fact that temperature gradients can accumulate alarmingly high temperatures at the heat sources (typically semiconductor junctions with limits no higher than 125 deg C in most cases) if care isn't taken to minimise the intermediate thermal gradients.
I marvel at how unventilated smpsu wallwarts manage to cope using conduction via internal air to pass heat to a plastic thermal conductor to the outside surface actually exposed to the cooling room air which is why I have so much confidence in their longevity when I salvage their innards for use inside of a ventilated case as an auxilliary psu such as the half amp 12v smpsu board I'm using to power the OCXO in my FY6600 generator (a grand total addition of another 1.3W to the heating load within the box).
It's all too easy to overlook the importance of effective cooling, especially when adding such items as replacement analogue PSUs with their own critical cooling requirements to an existing box full of electronic components regardless of all the empty unused space within just crying out to be filled with all that analogue psu goodness.
Many have fallen foul of the 7805/7812/7912 regulator's thermal requirements but I see, now that I've googled "OKI 78SR", that you've discovered a rather neat solution (90.5% efficiency!).
JOOI, have you measured the ripple noise output of this dc-dc converter substitute for the classic 7805 yet? I've got a similar dc-dc converter module on order from BangGood which claims less than 30mVpp ripple which I'm hoping to use in my very first basic GPSDO. It was part of an order totalling well below the 25 quid point at which you get free expedited delivery so is on a slow boat from China meaning it may be a few weeks before I get the chance to test its ripple noise performance. I've also got a step variable voltage dc-dc module in that order which made no mention of ripple noise performance which I'd also like to test.
I rather suspect that despite some level of ripple noise in these dc-dc modules, they'd be a good compromise between the high ripple noise of an efficient mains voltage smpsu and the lower noise of an analogue R type mains transformer psu based on these converters in place of the classic 7805 type regulator.
At the very least, such a solution should eliminate the direct radiation from the high voltage switching in a conventional cheap mains voltage smpsu from polluting the analogue circuitry of the signal generator either directly or via common mode conduction to the main board's low level voltage connections. The ripple noise of a dc-dc converter module should be easier to filter out than any common mode noise injected by a high voltage switching chip. However, whether any of this has any basis in fact remains to be seen but I do have high hopes that this turns out to be case.
As for the need to modify any of these signal generators, it does, as you point out, rather depend on the use you plan to put them to. For audio use, they'll be fine as they are, especially if you invest in a decent external 50 ohm 20dB attenuator with perhaps a further selection of attenuation steps allowing you to avoid the internal 86 ohm attenuator pad for voltages below the 500mVpp mark, eliminating the risk of noise from the psu leaking past this built in attenuator onto the BNC outputs.
Although audio signals are typically carried on notionally 600 ohm circuits (balanced or unbalanced) or simply passed from a low impedance source into a high impedance sink, it's still desirable to stick with a 50 ohm attenuator pad, if only for the sake of consistency. If impedance matching is important, a simple resistive matching pad can take care of that at the volt level since there's plenty of surplus voltage swing available to compensate the additional loss of such a matching pad.
As I see it, all of these signal generators will benefit from the addition of a cooling fan and the suppression of the half live mains voltage leakage with a 1 to 10K "static drain" resistor connection to the protective earth on a three pole mains socket, neatly avoiding the ground loop issue. If nothing else is required, these two modifications are, imo, essential prerequisites for a long service life and protection from the ESD risk posed to any devices under test.
A virgin FY6600 requires the most effort to complete these two mods, the FY6800 is already blessed with the required three pole mains socket for the leakage suppression mod leaving just the cutting out of a hole to accommodate a cooling fan whilst the FY6900 needs no such gross mechanical modifications, blessed as it is with both a C14 mains socket and an unpopulated fan aperture on its rear panel which I guess was just an incidental feature of the case Feeltech had bought in (I rather doubt they specifically asked for it).
JBG