Thanks Johnny for your long reply!
I read all you postings regarding the Feeltech FY series and the idea for my accomplished and planned modifications came from your
postings!
Some might consider you a 'glutton for punishment' for admitting that.
And sorry for my failure/error/mistake. I'm not a native speaker and I have to help myself with my east german school english!
No need to apologise! Your written English is far better than my written Spanish (Español) and I managed to obtain a CSE (Certificate of Secondary Education) grade 1 pass (regarded as the equivalent of a grade 4 pass in the higher GCE (General Certificate of Education) exam standard of the late '60s here in the UK).
I sometimes spot similar errors when I proof read these missives of mine (and I'm a native English speaker!!!
). I wasn't being critical, just contributing a bit of feedback to help you perfect your written English skills.
(and by the way: cool nick name with respect to a great song - I especially like it in the version of Jimi H.)
It's a pseudonym I came up with when I started posting to usenet news groups. Not my first (ill considered) choices of pseudonym but my 2nd or 3rd by which time I'd decided to use one that at least would be pronounceable when read out loud (and more easily remembered) with a loose connection to my actual name (John).
Many of the pseudonyms seen in EEVBlog postings seem to be based on electronic components without much regard as to how awkward and clumsy they might be for their respondents to use as a means of address in any replies. I know a name, real or pseudonym, is just a label which can be any unique combination of letters and digits. Since we're human rather than machines relying upon simplistic labelling algorithms where any jumble of letters and digits will suffice to uniquely specify a label, I'm left wondering at this predilection for machine friendly labels instead of the more easy for humans to remember labels such as Johnny, Frederic, Alfred, Alphonsus Liguori, Cuthbert and suchlike?
Unfortunately the bought OXCO has no Ref.Volt.Out and no adjustment (the pins are still there) - see attached data sheet of
the vectron OC-160-DAB-308BF. I wasn't aware of that but for my first tries it works out for me. Later I can go ahead with a better one.
But I'm impressed how carefully you watched my pictures.
You've fallen into the trap that OCXO
always means VCOCXO when the reality is that it only
typically means VCOCXO and can, as you've just discovered, sometimes just mean literally "What it says on the tin.".
Strictly speaking, what you had been
shopping for, was a VCOCXO but when you try using that as a search term, you end up missing most of the VCOCXOs on offer simply through them being mislabelled as OCXOs for the sake of descriptive brevity under the quite reasonable assumption that an "OCXO" without such a VFC pin is about as much use as an armless Handyman (i.e. Pointless!).
TBH, on a package that size which includes the most expensive part of the VFC circuit, the pin itself, the startling omission of the varactor diode and smd resistor network is, imo, sheer criminal negligence, no doubt perpetrated at the behest of the manufacturers' bean counters looking to shave a cent off the cost off a 50 dollar part.
It's rather unfortunate that such ambiguity from the common use of the label OCXO to describe both OCXO and VCOCXO types can lead to such purchasing errors when the buyer fails to check this very point. When purchasing from major electronic component suppliers such as Farnell and Digikey which list OCXOs by the thousand, it's easy enough to check whether you're looking at a fixed or tunable type (once you've found the appropriate data column in a page and half wide table!).
If you're looking at an Ebay offering and it doesn't explicitly state which type it is, you'll need to message the seller and ask them to clarify this point. That way, you have it in writing should the part supplied not be as they described, allowing you to extract a refund either directly or via Ebay's money back guarantee.
As for the non-functioning Vref pin, whilst a bit of a disappointment, it's far less of an issue. My very first ever VCOCXO which I bought last year at the Blackpool Mobile Radioham rally for the princely sum of 4 quid had a Vref and a VCF pin. It was a 13MHz CQE unit rather than the 10MHz VCOCXO I'd actually wanted to purchase. I bought it anyway since it had been the
only OCXO of any sort I'd seen anywhere on sale at that event and any OCXO was better than nothing, especially at such a low price. It would at least give me a VCOCXO to experiment with whilst I tried to track down a more suitable one at a less than exorbitant price on Ebay and I had a strong hunch that I could use it to generate a 10MHz reference locked to its 13MHz output with just the right mix of TTL magic if needs be (in this case, I wasn't wrong - the proof sits atop of my component drawers stack on my workbench).
I mention this DIY GPSDO project related trivia simply because I couldn't track down a data sheet to identify its pin outs nor its operating voltage (5 or 12 volts???) other than for a 12v 13MHz Vectron which helped for the pin outs but left me in the dark over its required operating voltage. However, this search for a datasheet did lead me to a UK based Ebay trader selling the 12v 10MHz version of this 13MHz CQE OCXO for just 99 pence more than I'd paid for that radioham rally bargain. I chanced 19 quid on three (£14.97 plus £4 P&P) which all tested good when they arrived just a few days later.
I then immediately placed an order for four more. At a fiver each, one can never have too many, plus bulk ordering saved a little on the P&P costs -2 quid for the first one and a quid postage for each additional unit - in this case, a total order cost of £24.95. These also tested good and I now have my "Lifetime's Supply" of CQE 12v 10MHz VCOCXOs.
Meanwhile, I had already lashed up a divide by 13 test circuit on a solderless breadboard sandwiched between times two and times 5 clock multiplier stages to prove I could actually generate a precise 10MHz from the original 13MHz OCXO which I had decided to treat as a 5 volt part rather than risk literally blowing my 4 quid investment with a dose of 12 volts.
I only had a relatively crude means of applying test voltages above the 5 volt mark to test with at that time, plus it was my one and only working OCXO and since it seemed to function ok on a 5 volt supply anyway, even if it was taking around 8 minutes to warm up, I decided to play safe and assume it was actually a 5 volt part after all rather than risk going beyond the 7.6 volts I'd already taken a chance with.
The way the square wave voltage had been increasing with supply voltage, some 2v p-p shy of the supply rail voltage, implying a never seen by me in any OCXO datasheets of a p-p voltage of 10v on a 12 volt supply, did strongly suggest that this 13MHz unit was in reality a 5 rather than a 12 volt part best not subjected to any further voltage abuse. I have, and still do, treat it as a 5 volt part to this day.
However, I'm working on a MK II GPSDO based on one of the 10MHz units and a simplified construction on copper clad board using 'squashed bug' style IC mounting to replace the current setup built onto veroboard (strip board) using the 13MHz OCXO with its associated collection of power rail polluting TTL to generate a rather jittery 10MHz reference which had required the addition of a 10MHz XTAL between the LPF and the output socket to tone down the jitter to a more acceptable level. This MK II version should be a vast improvement over its predecessor in spite of it being essentially the same basic analogue filtered hardware PLLed design.
Once I have this later version up and running, I'll be pulling that 13MHz OCXO out once I've collected enough test data comparing the old and the new units to run more supply voltage tests (to destruction if needs be) now that I have a suitable variable voltage bench supply to test with. I've had a growing suspicion these past few months that my "5 Volt" OCXO may actually be a cunningly disguised "12 Volt" part after all. Now that I have a 12v 10MHz OCXO, along with half a dozen working spares, I can now well afford the risk of blowing my 4 quid investment away in a 12 volt testing 'accident'.
To be honest: For my needs the FG works fine, I'm a hobbyist but I like good tools. So I bought some weeks ago the new Siglent SDS 2000X Plus
and with the "improvements" from the dedicated forum I own now a 500 MHz oscilloscope And with it's measurement function I discovered first
the freq-error of the FY. And there is a german saying: "Der Weg ist das Ziel" - which means that it's all about the way and not about the final result
Well, you don't exactly need to spend £2,736.00 on a SDS2000X Plus 350MHz 4CH DSO to measure that sort of frequency error
You could have made do with a £360.00 SDS 1202X-E (I paid a fiver more just over 18 months ago for mine) and a €20.00 u-blox NEO M8N to provide a cheap yet ever so precise (if a little jittery) 10MHz reference to compare against the FG's output (just tune the FG for 'zero beat' around a 180mH away from the 10MHz setting in your case).
The English expression is: "It's morel about the journey than the destination." That's ever so true of Life's journey with its unchanging itinerary: "You're born, you live, you die.". It would seem you're travelling First class whilst I'm happily making do with 'Economy Class'.
The frequency counter display, top right hand corner on these SDS models, is a hardware frequency counter display (6 digit for the SDS1202X-E and 7 digit for the SDS2000X Plus series) which is more accurate than the software based measurements display. When I first got my M8N to finally get a usable lock (the built in patch antenna had been demoted to an ornamental badge courtesy of all the circuit traces underneath its ground plane grossly overloading the tiny LNA with self inflicted RFI, effectively blocking all SV signal reception.
It was only when I tried a 3/4 wavelength wire antenna shoved into the SMA socket that I finally started seeing SV signals several days after I had taken delivery and several days prior to a separately ordered Active mag mount patch antenna with 5 metre cable being posted through our letterbox.
Once I was able to get the M8N module to lock onto satellites and allow me to program the PPS line to a 10MHz 50% duty cycle, I could finally check the accuracy of that SDS1202X-E DSO's hardware frequency counter for the very first time, discovering that it was some 300Hz high (10,0003MHz). This error reduced very slowly over the next 6 to 9 months, reaching the point where it now continues to give readings of 10,0000MHz on a 10MHz reference frequency.
I very much doubt that Siglent calibrated it with precisely the pre-aging offset required to let it settle "exactly on frequency" after 9 months to a year of service. They may possibly have given it some deliberate offset to allow for an average amount of ageing drift but the fact that mine, within the limits of the display resolution (6 digits' worth), is now 'spot on frequency' must surely just simply be a matter of pure dumb luck.
I would be a little leary of relying on a brand new Siglent DSO's hardware frequency counter to provide accurate frequency readouts even within its own modest 6 or 7 digit resolution limit. It's useful for showing 'ball park' readings by way of a quick sanity check but if you need better accuracy you'd do better to use the frequency counter function of an FY6600 with calibrated to better than 1ppb OCXO reference on a 100 second gate time.
I've got mine measuring the GPSDO's 10MHz +11.5dBm sine wave output (not connected to the external reference socket - the FG's OCXO is running free) on a 100s gating time and the final digit is alternating between 1 and 0. On average, it's over-reading by 5mHz but that might be something to do with it claiming to be detecting a 56% duty cycle.
This may or may not be true of the actual sine wave output from my GPSDO. The 10MHz square wave output coming from three of the 74HC14 hex Schmidt trigger inverters via 150 ohm resistors connected to the LPF filter via a 100nF capacitor to remove the DC component further filtered by a 10MHz crystal tuned to its series resonance in series with the output from the LPF I'd have thought should have well and truly removed any such duty cycle imbalance but I'm open to learning new and startling facts in regard of converting square waves into pure sine waves.
I may have missed a trick or possibly it's merely a defect in the FG's frequency counter input circuitry. It'll be interesting to repeat this test with the MK II GPSDO as my source of pure 10MHz sine wave test signal.
I'm coming as hobbyist from Audio electronics, renewal of old HiFi equipment as a REVOX Tape Recorder, amplifiers and such stuff. During my University time I
worked with digital electronics and Zilog Z80 in the 80s. Then I were for 20 y out of E-business and restarted as hobby some years ago
.
2 years ago I started with Arduino and played around with GPS. I have a lot respect of everything > 1 MHz and I try here to improve my knowledge.
Well, it's a similar story here. For a change, I won't bore you with the gory details suffice to say that FeelTech's bean counter led cheapskate cost cutting measures are like an old old friend. Even Akai have managed to perpetrate some staggeringly stupid bean counter induced tricks on their flagship tape decks (the cost cutting clipping distortion inducing trick used to save pennies on the dolby record and playback boards in a 500 quid GX630DB and the idiocy of using a rubber drive belt between the rubber faced tape counter idler to drive a separately mounted optical tape counter sensor rather than mount the sensor directly onto the idler wheel shaft saving both costs and improving accuracy (the very opposite of the usual bean counter trick
) feeding an electronic counter which could be set to show footage or hours minutes and second's worth of tape used or remaining on their even more flagship GX747 bi-directional record and playback 4 track stereo deck both of which had sideband scrape noise and wow and flutter figures well below the likes of Revox and their ilk.
I still have those tapedecks. The later GX747 (originally bought to replace my much modded but stolen GX630dB) got packed away into its original box barely used (no Dolby noise reduction) nearly forty years ago when I was reunited with my beloved (I'd invested a lot of energy in various modifications) GX630DB hardly the worse for its year or so spent in the hands of a "Fence". This is still sitting on top of one of my speakers gathering dust waiting for me to get round to resuming my 30 year old tape digitisation project. I may well get back to it one of these days but don't ask me when.
Regarding the PSU I'm with you. It makes no sense to follow all the suggested improvements without any idea what can be achieved.
In my head is a very simple approach. On my shelfs are dozens of old Laptop PSUs with enough power for everything but unfortunately only one voltage.
So I looked around and found TPS5430 based ready to use solutions at EBay (see attached photo)
The thing that puts me off using such laptop charging bricks is the possibility that they may introduce switching noise which could see any analogue or buck converter regulators as an easy open pathway onto the 12v rails. They might not be as big a problem as I fear but be prepared to test for any such possible 'surprises'.
The main issue with mains voltage smpsus is this business of switching a 170 to 340vdc HT supply to use a ferrite transformer to step down the voltage and galvanically isolate the low voltage output from the incoming mains supply. Unless this transformer has had an extra penny or two spent on adding a screening foil between the mains side windings and the low voltage output winding(s), a lot of high frequency switching energy will be coupled via the interwinding capacitance from the high voltage pulses onto the low voltage secondary windings producing a significant amount of common mode interference on the LVDC output wiring.
This is certainly the main issue with the psu board used by FeelTech in their function generators (that troublesome Y cap is intended to divert these common mode currents straight back to the source where they belong. Unfortunately for us humans and any sensitive DUT, this source is a high voltage DC modulated with the mains voltage waveform. The cap filters the DC component but this leaves us with a very high impedance (1.6M ohm for a 1nF Ycap) half mains 'touch' voltage. An irritating annoyance for us humans but an ESD risk to any sensitive DUT not already cross bonded to the BNC ground prior to connecting to its test point. That 5.6K grounding resistor neatly shunts this touch voltage to less than half a volt whilst neatly avoiding the creation of a low impedance mains earthing loop.
I checked out the TI datasheet on the TPS5430 you mentioned. It's actually (primarily) a buck regulator chip. Obviously the circuitry around the negative rail chip has been rearranged to provide an inverted boost output topology, sacrificing some efficiency and output capability and the inevitable increase in ripple voltage on its output. Provided it can still supply the demand from the load, the increased ripple can be mitigated with an additional LPF.
The problem with almost all boost inverters is that they have a harder task to perform than its buck converter counterpart. It's a problem that's made worse in the case of inverting boost converters which can't cheat by using the incoming supply as a base voltage upon which to series output a lower voltage to effect the required voltage boost. In the case of an inverting boost converter which supplies a negative rail from a positive supply, all of the output energy has to come from the energy storage element, typically an inductor.
The main reason for my considering the slightly less efficient R_Core transformer compared to its toroidal transformer counterpart is due to the lack of any toroids with more than two isolated secondary windings compared to the typical maximum of four typically found on R-Core transformers. The need to choose an efficient compact 25 to 30VA rated mains transformer is driven by the lack of cheap negative rail buck converter modules. Having at least three isolated secondary windings available will allow me to use cheap high efficiency positive rail buck converters for all three supply rails (using one 'connected backwards' for the -13.5 volt rail).
However, the 1.2MHz switching buck converters I'm favouring for this job only have a maximum input voltage limit of 24vdc which makes secondary voltage selection very tight when you don't know the exact basis for their 15 or 18 volt specification (on load? off load? at maximum or minimum mains voltage? What???). I'd be much happier with a version that is rated for a 40vdc input since it would ease my choice of transformer considerably. The only buck converters I have with a 42vdc input rating are the slightly less efficient switching/LDO combination type where the buck converter supplies 6.5vdc to a 5v LDO. They're supposed to be a compromise between the efficiency of switching technology and the low noise feature of an analogue regulator - Yeah, right! I'll believe that when I start testing them and see this with my own eyes
What do you think about this: using an external 19V Laptop PSU (no problems with grounding and security anymore), inside a TPS5430-based PSU
adopted to 2x13,5 V and another cheap Step-down-converter for 5V?? It's the cheapest way to (hopefully) improve the PSU of the FY. If there is enough
place and need for further "cleaning" you can bring another analog regulation behind the switching one. And I get enough power for any OCXO as well.
I played around with such a solution for a headphone amplifier which needs +/+16 volts and it works very well.
If you're going to be using DC charging brick limited to an output voltage of 19 to 23 volt, those cheap mini 360 1.3A variable output voltage buck converters sold by Banggood will be just the ticket for the positive rails, leaving you to track down just one positive input, negative output boost converter for the negative rail (otherwise just use that TPS5430 module for the +/-13.5 volt rails and a single mini 360 buck converter for the 5v rail.
Although the 5v logic supply won't be bothered by supply ripple voltage it may still be of benefit to add an lpf at the psu end to reduce any ripple since it's the effect of the corresponding ripple current flowing through a common return shared with the analogue supply that you need to keep mindful of.
If you can't separate the 5v logic supply's ground return from the analogue supply rails' ground return in the PSU, then aim to keep the ground connection between the PSU and the main board as short and as heavy gauge as is practical to minimise any such ripple current induced crosstalk between the logic and analogue supplies as possible. You can fit a ferrite ring or tube to suppress any common mode over the whole bundle of wires and even fit ferrite beads or tubes over the individual supply wires, just don't do the same for the common ground return.
But my main restriction is time. I'm still working (more than full time and with a lot of travelling) and have family. So I have to look for some hours to come forward with my projects.
I know what you mean, I was planning on making a start on my MK II GPSDO today but got so involved in creating this "Masterpiece" of a response
... Sorry, oh yes. here I am putting the final touch to this missive, which is to say, it's only a quarter to one in the am, still early for this night owl so I may yet despoil that lovely blank sheet of copper clad board with the very first hole to locate my chosen OCXO's final resting place in this new GPSDO project of mine before retiring for the night.
P.S. I may be slow in responding to any more postings from now on once I actually start building a new project as I'm hoping to do very shortly. Apologies in advance if I appear to be ignoring anyone but I'm not as good at this "Multi-tasking" malarky as I once had the arrogance to think I was.
P.P.S. I've attached a couple of documents you might find of interest.
JBG