MJ Statistical Regulator PCBs

[jack]

Over 20 kits now. Going to have to order more boards n bits....

[Ray P]

Just the one kit for me please Nick.

Ray

[jack]

Will build the test boards this weekend as the MKP caps have just arrived - nice bright red!

Assuming all goes well, first kits should go out in a week - I have had to order more of the caps as they are the tricky one to source and have a 2 week lead time.

Need to write some documentation (hate doing that) for the kits too...

I also want to build an HV uVolt interface to my audio analyser to measure the resultant noise from various configurations.

[jack]

If I never see another zener, it'll be too soon

These boards are a bit fiddly to assemble. Need very fine solder and a fine tipped, small, soldering iron... there's 102 solder points on a fully populated board...

[Cressy Snr]

If I never see another zener, it'll be too soon

Now you know how I felt
When my board failed to work the first time I tried it, after all that work,
I was ready to kill, believe me.

[Mike H]

This is why when I designed boards I thought about what's the actual real world dimensions of these component placements? And if necessary created new footprint symbols accordingly, as it's difficult the translate what you see on the screen into reality. For example make all holes 1mm (as per Veroboard) which in turn dictates a more generous sized pad that you can get a soldering iron onto more easily.

[jack]

You can specify all the pad dimensions etc. in Eagle's design rules - every aspect is adjustable - stringing, whatever...

Its more that my eyesight isn't what it once was and there are 40 of those tiny zeners to solder - just fiddly, nothing more.

Because I design mostly using surface mount, I tend to make my boards rather compact, perhaps unnecessarily so. However, the bottom line is that this board needed 40 zeners - no way out of that The boards are quite cute, though !

I'll test it tomorrow - just watching Guy Martin in India at the moment - great, innocent, fun

[jack]

My creation lives !

With about 260VDC in and a fully populated board (nominal 224V) I set the Izener to 20mA and using a 16K 50W load (15mA into that), all seems fine.

T1 runs at about 75C and and the zeners at about 85C which is expected - the leads of the zeners are closer to 100C as they reflect the junction temperature rather better, but it all seems very stable... Due to the zeners warming up and other effects, Vout is actually 237V rather than 224V - this is normal

Zeners are dissipating about 4.8W and T1, which has the full 35mA thorough it, about 1W in this scenario.

Left meter is Zener current, right is Vout

General test setup showing dummy load and Vin PSU in top left...

[jack]

Occasionally I get a bit carried away by Maths - I like to understand why stuff behaves the way it does. Specifically, why I was measuring 237V rather than 224V across the string of zeners (see post above).

For instance, in this thread we are discussing/playing with a circuit with a lot of zeners. You would expect with 40 x 5.6V (nominal) to get pretty close to 224V on the end of the string, because even though the datasheet says that their zener voltage can be anywhere from 5.2V to 6.0V, with 40 of them you'd hope that they'd average out to 5.6V? Wouldn't you?

If zeners were perfect devices and followed the CVD (constant voltage drop) model, when subject to a reverse voltage, Vr, no current would flow until Vr > Vz where Vz is the nominal zener breakdown voltage. At this point, as Iz (the reverse current through the diode) increases, Vz stays exactly the same, in this case at 5.6V.

However, the world is not perfect - both Tj (junction temperature) and Iz have an effect on Vz, and when you have 40 zeners in series, these effects can become significant.

In practice, a different model is used, the ZPWL (Zener Piece-Wise Linear) model.

The ZPWL model allows for a dynamic impedance that is in series with the diode - in the datasheets this is called Zz and is typically quoted at Iz = 5mA and Tamb = 25C. In the case of a BZX55C5V6, this is 25ohms (known as Zzt, the dynamic impedance at the test current, Izt).

So, you can see that as Iz increases, Vz will also increase due to the dynamic impedance.

But wait! It's a dynamic impedance - it's not constant, it changes with Iz, so we have to compensate...

The formula for how much the zener voltage changes (dVz) with changing Iz is:

Code: Select all

dVz = 2 * (Izt * Zxt) * (Iz - Izt) / (Iz + Izt)

Note: This formula is a good second-order approximation - its good, but not perfect

From the data sheet, Izt = 5mA and Zzt = 25 ohms. We need to find out how much Vz will change when we apply Iz = 20mA rather than 5mA.

Substituting back into the above equation we end up with dVz = 0.15V - not inconsiderable, especially when we have 40 of them. Remember this value for later...

Now we get to the thermal effects. Datasheet characteristics are quoted at Tamb (ambient temperature) of 25C (typically) - the zener's Vz will change as the temperature moves away from Tamb - this value is quoted normally as a %/C, i.e. the percentage change in Vz for every degree C you move above Tamb - for zeners above Vz = 5V this is normally a +ve value, below 5V its typically -ve.

In the case of the BZX55C5V6 this can vary between -0.01 & +0.06 %/C, which averages out at 0.025 %/C.

At a Tj of about 90C, we are 65C above ambient, so the thermal change in Vz will be 65 * 0.025 * 5.6 / 100 = 0.091V

Adding both the thermal and ZPWL corrections together, we get a final Vz under our operating conditions of 5.6 + 0.15 + 0.09 = 5.84V

So, our real Vz should be around 5.84V, therefore with 40 of those, we end up with about 234 volts rather than the simplistic 224 volts we might have thought.

I measured the output at 237V which is only about a 1% error.

Hope that clears up a few things about the behaviour of zeners!

[Mike H]

Yes as the other Nick once reminded me, Zener Voltage increases with temperature.

[jack]

Yes as the other Nick once reminded me, Zener Voltage increases with temperature.

Not really - for low zener values, Vz drops with increasing temperature; for higher values (above about 5V), it increases, but only slightly. See the attached datasheet...

The temperature effect, as demonstrated above, is significantly less than that due to increasing Iz - you need to consider both if you are at all concerned with accuracy...

BZX55C.pdf

BZX55C-series datasheet from Farnell

(24.9 KiB) Downloaded 248 times

[jack]

Here's a small spreadsheet I knocked up that does the maths for you - you can enter the datasheet values for the zener of your choice (it's preset for a BZX55C5V6) and the zener current you want (between 10 & 20mA) and it'll tell what the real likely value of your zener will be - note that this really only works for TRUE zeners, i.e. below about 6 volts. Does all the maths for you... main use is calculating the required actual zener string length you are likely to need, plus allowing you to see the effect of environmental changes or the effect of not using an average value of Tcvz or Vz etc.

EDIT: OH! No Excel files !

Email me if you want a copy....

Have to be satisfied with a screengrab !

[jack]

Apologies for the delay in getting these kits out - the status is that everything seems to work fine but I want to run some longer soak tests, however I have pretty much no free time at the moment as the company I work for has just been taken over (see "Nothing in particular" for a whinge by me about all that...)

Another week or two and it should be fine

[jack]

Just managed to get some "lab time" (*).

At Iz = 10mA, Tj drops to about 55C and the boards seems very very stable. Noise floor is now below what I can measure - MJ recommends a minimum of 10mA, but noise rolls off even more up to 20mA,,, Downside is that at higher current, not surprisingly, the zeners get a lot hotter...

The spreadsheet calculates an adjusted 5.725V Vz under these conditions making 229V adjusted vs. 224V nominal for a chain of 40 - I measured this at 233V which is a 1.7% error - pretty good, IMHO - with Iz = 20mA this drops to 1.4%.

If possible, I'd like a couple of volunteers to build & test, i.e. ACTUALLY USE, these boards... If I'm supplying all the bits, it'd be only fair to ask for some feedback before I order several 100 quids' worth of stuff

Any volunteers?

I'll supply all the bits - I just want you to build, use & listen to the results and then report back either way as to what, if any, the effect on the SOUND was....

Ta very much

(*) never easy - spent last night in A&E with #1 son who dislocated hip & damaged hamstring during a league rugby match...

[Paul Barker]

As my amp is a long way off I can't volunteer.

Good idea though.