FeelTech FY3200 PSU upgrade

[leonerd]

I've got one of these popular DDS signal gens, the FT3200 - mine's the full 24MHz unit. As a lot of people have discovered, the switching power supply inside it suffers a lot of mains leakage. I measured about 95Vac between its "ground" outputs and earth mains, and it seems capable of supplying nearly 1mA of current!

I've therefore decided to undertake a little project to upgrade the power supply, on a similar project to

 

I measured the current requirements of the lines - a little more directly than the above video (watch from around the 2m40 point onwards), but I get very similar results, namely:

  +5V - current draw increases with output frequency, independent of output voltage or load. Maximum occurs at 24MHz, of 260mA.

  +12V and -12V - current draw increases with output current, independent of frequency. Maximum occurs at 20Vpk into 50ohm, of 80mA on each.

I'm thinking rather than 3 linear supplies, what I'll do is make 2 linear supplies for the +/-12V lines (as they power the analog output frontend), but go with a simpler efficient switching design for the 5V rail. In particular, most of the logic chips on the board are 3.3V anyway. The 5V line into the actual board mostly just powers a bank of 3.3V regulators, and possibly the LED backlight of the display. I think that will be fine with some switching noise.

I expect I'll post a few photos and diagrams and the like on this thread. Wish me luck...

[Kevin.D]

I've got one of these popular DDS signal gens, the FT3200 - mine's the full 24MHz unit. As a lot of people have discovered, the switching power supply inside it suffers a lot of mains leakage. I measured about 95Vac between its "ground" outputs and earth mains, and it seems capable of supplying nearly 1mA of current!

Apart from being enough to fry sensitive components every time you connected
some node of your own circuit to earth this also would not pass any electrical safety laws in
Europe (or most of the world I think).
Since this is a class II device (accessible unearthed conductors) max leakage should be .25 mA

Copy/Paste from Standard IEC950 entitled “Safety of Information Technology
Equipment Including Electrical Business Equipment"

Earth Leakage Current: Maximum earth leakage
current must not exceed the limits shown in the following table
under the most unfavorable (highest) input voltage. For class II
equipment when output is not connected to earth ground, the
test shall be made on accessible conductive parts, and to metal
foil with an area not exceeding 10 cm x 20 cm on accessible nonconductive
parts.

Class       Type of Equipment            Maximum Leakage Current
II                  All                                        0.25mA         
I                Hand-held                                 0.75mA
I                 Movable                                   3.5 mA
          (other than hand-held) 

[leonerd]

Some back-of-envelope calculations... Anyone see anything wrong with the logic here?

If I go with a typical LM317/LM337 linear regulator combination for the 12V lines they'll want 3V of dropout on each rail for the regulators. Add a bit for rectifiers, account for RMS vs. peak at input-side smoothing cap, and I think I can use a 15/0/15V[rms] secondary winding on the transformer. I need 80mA at 15V which means 1.2VA, and again for the negative rail, is 2.4VA required by the linear regulators.

If I tap off from the input to the +ve regulator and ground to supply the LM2597 board to generate 5V, I'll need more power here. To generate 5V at 250mA in a buck regulator that I'd guess is around 80% efficient needs 1VA.

So in total I need at least 3.4VA of power from the transformer.

Is 4VA going to be sufficient, or might I want more spare capacity? Are there any downsides to overspecifying the transformer and getting a 6VA (or more) model?

Is a 15/0/15V going to be sufficient for these regulators? 15V[rms] is 21.2Vpeak which is 20.2Vpeak after the rectifier (1 diode drop because it's a split rail), so even letting the smoothing capacitor droop to 80% of that still gives us 16.1V into the regulator, giving it a helpful 4.1V of headroom to generate the 12V output. So I think that sounds fine.

And finally, is the transformer/rectifier going to be OK with me asymmetrically loading it? The LM317 and LM2597 between them will be pulling about twice the current from the +ve end of the rectifier as the LM337 will be sinking into the -ve.

[StillTrying]

If these are sustained currents I think you're missing a VA or two and 4VA won't be enough.

I need 80mA at 15V which means 1.2VA

The 80ma is drawn from a ~20V source so 1.6VA.

To generate 5V at 250mA in a buck regulator that I'd guess is around 80% efficient needs 1VA.

It must be at least 1.5VA. @ 20V.

And finally, is the transformer/rectifier going to be OK with me asymmetrically loading it?

Yes, you're not really asymmetrically loading the transformer at all.

[leonerd]

If these are sustained currents I think you're missing a VA or two and 4VA won't be enough.

I need 80mA at 15V which means 1.2VA

The 80ma is drawn from a ~20V source so 1.6VA.

Oh, because of peak vs. RMS?

To generate 5V at 250mA in a buck regulator that I'd guess is around 80% efficient needs 1VA.

It must be at least 1.5VA. @ 20V.

Righty. So 3.2VA for the linears and another 1.5VA for the buck; that's 4.7VA. Turns out that 6VA is about the smallest transformer anyone sensibly sells, so that should suffice, yes?

And finally, is the transformer/rectifier going to be OK with me asymmetrically loading it?

Yes, you're not really asymmetrically loading the transformer at all.

Well, not the transformer, but definitely the rectifier. I'll be putting twice as much current through the two diodes generating the +ve output, as I am the -ve.

[StillTrying]

I'd be tempted to try a 12-0-12  @ 10VA but many would disagree with that!

[leonerd]

I'd be tempted to try a 12-0-12  @ 10VA but many would disagree with that!

12Vrms is 16.97Vpeak; minus 1V of rectifier drop that's 15.97Vpeak into the regulators. The regulators want 3V of dropout to generate 12V, so that's nearly 1V spare headroom. This is just within possible, but means the smoothing caps can droop by no more than 6%.

On the plus side it does mean less work for the linear and buck regulators to do, so less waste heat generated. Good inside a passively cooled box.

[leonerd]

12Vrms is 16.97Vpeak; minus 1V of rectifier drop that's 15.97Vpeak into the regulators. The regulators want 3V of dropout to generate 12V, so that's nearly 1V spare headroom. This is just within possible, but means the smoothing caps can droop by no more than 6%.

To achieve an acceptable ripple with the load on the +ve cap of 160mA, we're looking at ballpark of over 3000uF to smooth that. Whereas a 15/0/15 transformer works just fine with a little under 1000uF.

The regulator board I'm looking at (eBay link above) comes with 2200uF caps.

[CJay]

12Vrms is 16.97Vpeak; minus 1V of rectifier drop that's 15.97Vpeak into the regulators. The regulators want 3V of dropout to generate 12V, so that's nearly 1V spare headroom. This is just within possible, but means the smoothing caps can droop by no more than 6%.

To achieve an acceptable ripple with the load on the +ve cap of 160mA, we're looking at ballpark of over 3000uF to smooth that. Whereas a 15/0/15 transformer works just fine with a little under 1000uF.

The regulator board I'm looking at (eBay link above) comes with 2200uF caps.

I seem to remember the PSU from an old ADSL router I used to have dozens of would do for that job, +5 -12 and -12 all in one standard size open frame module.

I've got a small pile of them sat in my shed if you want to try one (and if my memory is correct on the -12 rail being present)

[leonerd]

I seem to remember the PSU from an old ADSL router I used to have dozens of would do for that job, +5 -12 and -12 all in one standard size open frame module.

I've got a small pile of them sat in my shed if you want to try one (and if my memory is correct on the -12 rail being present)

Well that is what I'm aiming at. The question is whether the +12/-12V lines are linear or switching though. My intended design here has those supplied by linear regulators, because they're the analog output chain. Also it leaves me with a simple two-pole transformer primary of the kind that would let the signal generator float relative to mains earth, or at the very least not let through that nearly-1mA of mains noise onto the output ground.

[CJay]

I seem to remember the PSU from an old ADSL router I used to have dozens of would do for that job, +5 -12 and -12 all in one standard size open frame module.

I've got a small pile of them sat in my shed if you want to try one (and if my memory is correct on the -12 rail being present)

Well that is what I'm aiming at. The question is whether the +12/-12V lines are linear or switching though. My intended design here has those supplied by linear regulators, because they're the analog output chain. Also it leaves me with a simple two-pole transformer primary of the kind that would let the signal generator float relative to mains earth, or at the very least not let through that nearly-1mA of mains noise onto the output ground.

For the +12, definitely from the main transformer and I *think* the -12 will be as well but I wouldn't swear to that.

They're  a pretty quiet supply though, ADSL doesn't play nice in the presence of high frequency noise.

[leonerd]

Status update: I ordered an adjustable LM317/LM337 regulator board (http://www.ebay.co.uk/itm/262483314472) and a 2x15V 6VA transformer (http://www.ebay.co.uk/itm/121671766834).

I'll post again when the bits have arrived and I've had a chance to build something up.

[leonerd]

  +12V and -12V - current draw increases with output current, independent of frequency. Maximum occurs at 20Vpk into 50ohm, of 80mA on each.

Waaaait a second. 20Vpk-to-pk is a little over 7Vrms. Into a 50ohm load I'd expect to see 140mA. And yet I see only 80. That doesn't add up somehow.

Or am I somehow a factor-of-two off here because of two rails?

[taemun]

  +12V and -12V - current draw increases with output current, independent of frequency. Maximum occurs at 20Vpk into 50ohm, of 80mA on each.

Waaaait a second. 20Vpk-to-pk is a little over 7Vrms. Into a 50ohm load I'd expect to see 140mA. And yet I see only 80. That doesn't add up somehow.

Or am I somehow a factor-of-two off here because of two rails?
[/quote]

There'll be a 50 ohm resistor in series with that, inside the unit. So it's a 100 ohm load on a 7Vrms source - 70mA plus some spare change to get to 80mA.

[hansibull]

I got my Feeltech FYS3200 in the mail today, and decided to replace the PSU after measuring 17v AC between mains ground and the BNC connector. This is the result after a few hours of work:

[leonerd]

I got my Feeltech FYS3200 in the mail today, and decided to replace the PSU after measuring 17v AC between mains ground and the BNC connector. This is the result after a few hours of work:

Looking good. That's linear all the way there I see? Quite a large heatsink required there.

[masterx81]

I want to use a single dual 15+15v 5+5va output transformer, 7812 + 7912, and a buck for the 5v output, i'm in the process of making it...
What's the advantage of using the LM317/LM337 instead of the 7812/7912?

[hansibull]

Looking good. That's linear all the way there I see? Quite a large heatsink required there.

Yep, it's linear all the way. Haven't had any heat related issues yet, but I haven't stress tested it for several hours either

[leonerd]

What's the advantage of using the LM317/LM337 instead of the 7812/7912?

No reason at all. I just happened to find a nice-looking cheap regulator board on eBay that is based on the LM317/LM337, so I went with that. I'd have been just has happy to go with a 7x12 pair also.

[leonerd]

My PSU board eventually arrived from eBay, so I've been able to complete this.

I'm using a piece of black perspex as a base board to mount the LM317/LM337 module on, along with the transformer, and the 5V buck regulator to provide the digital section power. New mains wiring and the addition of a fuse and a nice chunky MOV to finish it off.

[technogeeky]

Very nice!

Do you have any idea what effect this has on the signals generated, if any? Or is this project just to get rid of the mains leakage?

-tg

[leonerd]

Do you have any idea what effect this has on the signals generated, if any? Or is this project just to get rid of the mains leakage?

Oh I didn't really have any problem with PSU switching noise on the outputs. I was purely trying to get rid of the mains leakage. I expect any nonlinearity out of the R-2R ladder (see various other posts on that subject; especially around the zero-crossing region) would outweigh any PSU noise at this point anyway.

[StillTrying]

Looks neat to me, I wouldn't have bothered with the MOV.
What was the smoothed voltage from the 15V TX, I still think a 12V would just be enough.

[Excoriator]

I bought a 1:1 isolation transformer and kept the switch mode supply. Mounting it was a doddle. I used brackets to spread the weight of the transformer over four bosses, but even including these brackets, the job took under an hour with minimal intellectual effort included.

It allows me to 'float' the unit provided I don't connect the USB socket to the computer. This is not a serious constraint as far as I'm concerned, but for those who find it is, you can buy USB isolation cables that solve that little problem too.

Chucking out the switcher for a linear PSU redesign is a bit OTT for me.

If you can't find a suitable 1:1 transformer at a suitably low cost, then two small 1:n transformers back to back would do much the same thing. The power consumption is tiny, so two dirt cheap transformers should cost you peanuts!

[S185149]

Very late to the game here. Can anyone tell me what type of connectors are used in the FY3200 for the power connections?

Thanks,
S18