1A2 KSU Revisions - what changed and why: REV J7

[Greg Ercolano]

In these "revision" threads I'll try to focus on particular EE design issues
I ran into over the years during development of the 1A2 KSU boards.

First one is the most recent change in REV J7, in November 2024.

A customer was making a custom device for which the 1A2 KSU board was just
one component of a larger system. In his design he needed to bypass the on-board
300Ω series resistor in the 105VAC ring voltage circuit, whose presence is recommended
by ring generator documentation to protect the supply from dead shorts across the
ringer conductors.

Normally I don't recommend changing things like this, but doing so revealed the
contact open/closures during ringing caused inductive voltage spikes from the
phone set ringers to create enough EMI to cause the onboard PIC chips to
intermittently reset, causing calls to fall out of Hold when other lines were ringing.

I was able to see this on the scope pretty clearly as large ~1KV spikes,
depending on where in the 105 VAC ring voltage the contacts opened:

The series resistor kinda hid the problem, making the Hold problem less likely
to occur, but could still happen once in a while.

I thought this was something definitely worth investigating and solving
to prevent calls falling out of Hold, no matter how infrequent the situation.

To solve, I tried various circuits and much research before settling on adding
bidirectional TVS diodes across the relay contacts, which clamps the voltage
spikes as soon as they cross the device's Stand-off Voltage (V_R) rating.

I later learned Western Electric had added this in later versions of their
line cards using time period equivalent devices (see below).

In my case spikes could occur on either side of the AC wave (as shown above),
and at 105 VAC (RMS), the AC wave ends up being about +/-150V peak
(in actual practice, ring generators vary in their AC output voltage and waveform
quality suffers in the presence of inductive ringers and long wire runs)

Since the diodes clamp at DC voltage ratings, I decided to leave at least 30V
of headroom around the 150V peaks, choosing a V_R of 185V, which in my case
ended up being the 1.5KE 220CA (from Littlefuse Inc). The ratings for this device:

• V_R (Standoff Voltage) 185V
• V_BR (Breakdown Voltage) Min=209V, Max=231V
• V_C (Max Clamping Voltage) = 328V
• Ipp (Max Peak Pulse Current) = 4.6A
  

The bidirectional TVS diodes symmetrically clamp on either side of the AC wave,
as shown in the right hand diagram (shaded in yellow):

On the scope it very clearly clamped the unwanted spikes very quickly and right
at the rating limit, as shown here:

At this rating the devices are necessarily beefy; about twice the size of the 1n4005
ring diodes (used for ring programming elsewhere on the board), e.g.

But they work great, are single package devices, low cost, fast-acting, and are
really designed for this exact purpose, clamping out inductive spikes on AC devices.

Some options I investigated before settling on the bidir TVS:

• Traditional RC based snubber circuits.
  Ruled out because the caps were physically large, and two components
  were needed per line. Also the RC snubbers didn't squelch the spikes as well
  as the TVS diodes.
  
  
• MOVs
  Ruled out because these are really meant to protect against destructive "single events"
  (like lightening), and slowly self destruct over time due to their construction; basically
  a lot of grains of conductors that only begin conducting when a large surge passes across
  them, and the grains basically weld shut, causing the device to eventually create a short.
  

The term TVS (Transient Voltage Suppression) seems to be the generic name,
but during my research I've also seen them referred to as "transorb", "transil",
1.5KE, and other names.

Bell System / Western Electric
Similar devices were used by the Bell System, calling this device a "bidirectional voltage
regulator diode", and I noticed these were added to their later KSU line card designs,
shown in circuit with this schematic symbol:

..that little "L" below the symbol is actually a "symbol" or shape, and not a letter.
Western Electric used a lot of weird symbols, and sometimes changed them over time.

In the above bell system schematic, that "L" is sometimes found flopped horizontally
in later designs, perhaps to make it clear it's not an "L". Pretty sure this is the same thing
as the above, as there's no other symbol like it, since all the text in the above schematic is
italic, whereas that "L" is not, perhaps to help differentiate from such symbols. (shrug)

It seems they changed this symbol over time:

(excerpted from a Western Electric draftsman's standard document on schematic symbols)

Apparently TVS diodes are not necessarily implemented as simple back-to-back zener dides,
as the schematic symbols might imply. They tend to be slightly more complicated in design,
as showin in this app note image a friend sent me:

(excerpted from: https://blog.semtech.com/tvs-its-just-a-diode-right-part-one)

[Greg Ercolano]

On 8/20/25 15:04, Greg Ercolano wrote:

most recent change in REV J7, in November 2024 [..]

I added bidirectional TVS diodes across the relay contacts, which clamps the voltage spikes

    Oh, and I should add, these new TVS diodes that prevent ring circuit voltage spikes
    were added to this section of the board:

    All recent orders have been getting this J7 revision board since this
    issue was identified.

    Older boards without this feature can still benefit from transient
    suppression if you connect one TVS diode across the S-Y and Y-S
    terminal screws inside each phone set extension that's programmed
    to ring, solve the problem at its source (the bell ringer coils). Doing
    this will prevent the issue of dropped held calls during ringing.

    If you have questions about this, feel free to follow up here
    in this thread.