Repair of the TDS 7x04 PSU Front End

NOTE: THE TDS 7X04 PSU EXPOSES POTENTIALLY LETHAL VOLTAGES. DO NOT ATTEMPT REPAIR UNLESS YOU HAVE

THE KNOWLEDGE, EXPERIENCE, COMFORT LEVEL, AND TOOLS TO WORK ON SUCH CIRCUITS SAFELY. NEVER WORK ON

A LIVE CIRCUIT ALONE. THIS NOTE PROBABLY CONTAINS ERRORS. ALWAYS CHECK ESSENTIAL FACTS PERSONALLY.

I recently purchased a dead TDS 7404 on eBay. The seller said that the unit was not working, and blew the mains fuses

when turned on. The unit was otherwise cosmetically in good shape. I powered up the PSU outside of the scope and

discovered that the seller was incorrect about the fuses; the PSU with no load tripped the 20A arc-fault circuit breaker

feeding the outlet into which it was plugged before the fuses had time to blow. There was also a momentary un-

localized arcing sound.

Tektronix was happy to quote $3800 for a new PSU. Neither Tektronix nor Martek (who made the custom PSU for

Tektronix) answered email inquiring about the availability of schematics. I was unable to locate schematics or further

useful details on the usual Tek scope web sites.

Preliminary Investigation:

The TDS 7404 service manual (pg. 6-66) provides a list of output voltages for this supply. Simple inspection tells us that

this is a multi-output, line voltage primary, switching power supply. The PCB has a white line delineating the high voltage

and low voltage parts of the PCB. There is a sticker on the side of C18 (one of the large, 470uf, 450V primary filter

capacitors) that provides typical current loads for each output. From those two sources, I was able to construct the

following table (note that +5V is produced and filtered separately for each connector, as opposed to +3.3V, which

appears on both connectors from the same source):

J2 V A Pins Source

+3.3 (see J1) A/B/C 1,3,5,7,9,11 T1

-15 3 A/B/C 13 T7

+15 3 A/B/C 15 T7

-5 15 A/B/C 17, 18, 20, 21,22 T7

+5 22 A/B/C 24,25,27,28,30,31 T7

J1

+12 6 A/B/C 5,6 T10

+5 18 A/B/C 8,9,11,12,14,15,17 T1

+3.3 35 A/B/C 19,21,23,25,27,29,31 T1

+9.8 (Vfan) 2.5 B/C 3 T10

Vcontrol

+5 1 N/A T10?

The label affixed to C18 also tells us that the design parameters for the PSU were 100-240VAC, 50-440Hz input voltage,

at 380 watts maximum.

The PSU actually consists of three multi-output switching supplies with a common front end. The “Source” column

above identifies which outputs derive from which supply. The PCB part number of the switching transformer of each of

the three supplies is shown in this column. “J1” and “J2” refer to the two 96-pin main connectors by which the PSU

provides power to the Acquisition Board (J2), the NLX Board (J1, via the PPC Board riser), and the PPC Board(J1).

The picture below identifies the essential elements of the PSU. The red circles identify which of the 1500uf, 35V

capacitors filter which output voltage. The three switching transformers and their PWM switching transistors are located

as shown. Each of the three subordinate power supplies is protected by an in-circuit ceramic fuse, with values as shown.

Since none of these fuses were blown, I concluded that my problem had to be in the front end. Unfortunately, in-circuit

testing (powered down) of the front end did not identify any shorted or otherwise failed components. In particular, the

bridge rectifier, which was a likely suspect since it sees both high inrush current and high-voltage inductive spikes,

appeared fine with an ohmmeter diode tester. Ditto for the front-end caps and semiconductors. I considered bringing up

the PSU on a variac until something bad happened, but I was concerned that internal arcing could create damaging

transients elsewhere in the front end. At that point, I needed to figure out how this thing actually works.

Theory of Operation:

Note: This section describes how I think things work. Feedback and corrections are welcome.

Input Filter - The PSU input supply voltage can range 100 to 240VAC, 50-440Hz. This means it can operate essentially

anywhere in the world without adjustment (other than fuse type). Using the simplified schematic below (created by me

from inspection; the inductance values are as measured), the line input passes through a simple canned RFI filter

module, then through fuses on both sides of the line (required if the line input is 220V), though the power switch (again,

both sides are switched), before reaching the PSU PCB through a two-pin AC-rated connector. The line voltage then

passes through a common-mode LC RFI filter network consisting of L1, L2, and L5; and C1, C2 and C69. Each of the

inductors consists of two separate but equal (same number of turns) windings sharing a high permeability toroidal

ferrite core. This makes it easier to get a lot of inductance with a small number of turns (important, since the wire size is

large for current-handling reasons). The anti-phase wiring arrangement ensures that the magnetic field resulting from

series-mode AC line current cancels to zero in the core (except for small leakage inductance). On the other hand, for

common-mode noise (high frequency currents or voltages that appear on both supply lines at the same time with

respect to ground); the two windings are in parallel and in phase. This presents a very high inductance between any

power supply noise source and the line input, so common-mode noise currents from the switching circuits in the power

supply will be bypassed to ground by C1 and C2 before they are passed to the input. C69 is a series-mode decoupling

capacitor; R1 ensures that any residual charge is dissipated when power is off, and perhaps helps a little to mitigate

large turn-on transients. I see no reason why R1 needs to be a 1% resistor.

Bridge Rectifier – The PSU uses a GBPC2506W (600V, 25A) full wave bridge rectifier, whose output voltage is 120 x 1.414

= 170V (for 120V) or 240 x 1.414 = 340V (for 240V). If replacement is indicated, a GBPC3508W (800V, 35A) provides

better protection from transients (IFSM is 400A vs. 300A). Capacitor C6 acts as a small reservoir in front of the boost PFC

circuit, preventing its input voltage from dropping to near zero when the input AC changes polarity. This limits any duty

ratio / switching frequency discontinuity. The design rule of thumb for this kind of capacitor is 3uF per KW, so with a

380-watt supply, 1uF is right on target. NTC Thermistor R7 has an initial resistance of 7 ohms, which helps to limit inrush

current on startup (when C18 and C19 are discharged), and then decreases in resistance to near zero as the thermistor

warms up.

Boost Power Factor Correction – R73, L3, CR15, Q3, Q6, C16, and CR9 (plus all of the circuitry in the green box) comprise

a boost-mode power factor correction circuit. This circuit is controlled by a TI (Unitrode) UC3854 power factor

preregulator, which uses average current-mode control to maintain sinusoidal line current, keeping the input power

constant with varying line voltage. Average current-mode control is a subset of Fixed Frequency Continuous Conduction

Mode (CCM) that uses current and voltage difference amplifiers, an analog multiplier and divider, and a fixed frequency

PWM to monitor the current in boost inductor L3 (kept continuous over the switching cycle), and make it track a

sinusoidal reference by selectively (and simultaneously) turning Q3 and Q6 on and off.

Details of operation for the UC3854 device are readily available online. See, for example,

http://www.ti.com/lit/ds/symlink/uc3854.pdf. In short, this circuit is a boost converter, whose output voltage must be

higher than the highest expected AC line voltage. Boost inductor L3 operates in “continuous mode,” meaning that the

duty cycle (when Q3 and Q6 are on) is dependent upon the ratio between input and output voltages. This duty cycle is

governed by four inputs to the UC3854: Vsense (the DC output voltage), IAC (the line voltage current waveform), Isense

(the line current, as determined by the voltage drop across R73, the current-sense resistor), and VRMS (the RMS value of

the line input). R73 is a 50 mOhm high wattage (its size suggests 5-10watts, but there are no wattage markings on the

part) current sense resistor on the boost inductor ground return path. Two 24N50 N-Channel MOSFETs (Q3 and Q6) are

paralleled to handle the current demand. If replacement is indicated, the 24A, 500V 24N50’s can be replaced with two

faster, pin-compatible 30A, 600V IXFQ30N60X’s, or conceivably with a single 60A, 500V IXFQ60N50P3. NOTE: These

parts (both the 24N50 and the IXFQ30N60X) electrically connect the drain and device heatsink tab. This means that

the PCB heatsink will be at HV DC ground potential unless measures are taken to electrically isolate Q3 and Q6. The

fact that C21 bypasses the PCB heatsink to HV DC ground suggests that the designer intended that Q3 and Q6 be

electrically isolated from the heatsink, although this was not done on my PSU. Q3 and Q6 share a heatsink with CR15,

the boost diode. CR15 (STTA12-060) is an obsolete 600V, 12A “ultrafast” recovery (55ns reverse recovery) diode. If

replacement is indicated, a Vishay FES16JT-E3/45 (600V, 16A, 50ns reverse recovery) looks like a good substitute. CR15

must be electrically isolated from the heatsink.

Bypass Diode CR9 (MR756; 600V; 6A) absorbs a significant portion of the inrush current during startup, helping to

reduce the transient load on boost diode CR15. The MR756 can be replaced with a Diodes Incorporated 10A07-T (1000V;

10A) if replacement is warranted. Safety capacitors C20 and C21 (4700pf; 250VAC) provide a path to ground for any

high-frequency noise on the high voltage DC ground. Bulk filter capacitors C18 and C19 (470uf, 450V) are paralleled to

provide a total of 940uF. These United (Nippon) Chem-Con KMH series capacitors are still available, but the newer LHS

series part (ELHS501VSN471MA50S; 470uf, 500V), offering both longer life and higher maximum voltage, are a better

alternative if replacement is warranted.

European high line voltage (255V * 1.414 = 361V) drives the design boost DC output voltage, which is likely to be near

400 VDC.

Switching PWM and Downstream Circuits - The 400 VDC output of the boost stage is passed through three in-circuit

fuses to the switching (chopper) transistors (not shown on the schematic above), which provide high frequency pulsed

DC (assumed to be around 100KHz) to the three switching transformers. Downstream of the three switching

transformers are voltage and current regulators for each voltage required (as well as thermal shutdown protection).

Since my PSU is not likely to have a problem in this area, explanation of these circuits is left to the intrepid reader. The

picture above identifies the output filter capacitor location for each output voltage.

Analysis: What’s Wrong with My PSU?

To trip the supply breaker, or blow a fuse, I would expect to see one of the following causes (probable causes of such a

short are shown in parentheses):

1) a line short to chassis ground (C2, C3, or C20),

2) a short across the AC line input voltage (R1, C69, or BR1, and (less likely), L5, L1, or L2), or

3) a short across the bridge rectifier output (C6, Q3, Q6, C16, C18, or C19.

There are other possibilities (e.g., a faulty UC3854 might be forcing Q3 and Q6 to stay on), but it seemed wise to start

with more likely and then move to less likely. Testing ensued. All of the components highlighted in green tested fine

when removed from the circuit and tested (at the relatively low voltages hand-held testers use). C18 and C19 both read

more than 20% low, so they moved to my to-be-replaced-regardless list. C16 came apart when being removed from the

board, so it was clearly a candidate for replacement (and might even be the source of the problem, although I saw no

signs of arcing). I was able to test L5, L1, and L2 simply plugging in the PSU with the other components removed.

Having found no smoking gun, and given their relatively low cost and the potential for hidden damage from arcing, I

decided to replace all of the above highlighted parts, except the inductors (which I knew were good at operating

voltage). I also replaced CR15 with a higher current faster part, as well as a few other diodes and capacitors exposed to

AC line voltage (and all electrolytic capacitors in the HV section), for the same reasons. The table below enumerates the

replaced parts and how to obtain them.

Qty PCB Part No. Orig. Part No. Orig. Value Replacement

Value Replacement DigiKey

Part No.

1 BR1 600V; 25A 800V; 35A GBPC3508W-ND

2 C2,C3 2200pf; 250V; safety cap same (300V) 490-9557-3-ND

1 C6 1uf; 450V; 10% 1uf; 630V; 10% 338-4132-ND

1 C16 0.033uf; 1KV; X7R same 490-8951-ND

2 C18/C19 470uf; 450V 470uf; 500V 565-5114-ND

2 C20,C21 4700pf; 250V; safety cap same (300V) 490-9563-1-ND

1 C69 .33uf; 250VAC; 10% .33; 275VAC; 10% 399-5971-ND

1 C93 6.8uf; 50V; 20% 6.8uf; 100V; 20% 493-11616-1-ND

4 C95,97,500,807 68uf; 35V; 20% 68uf; 50V; 20% 732-9596-1-ND

1 C808 .047uf; 630V; 10%; X2 same BC5120-ND

1 C900 330uf; 50V; 20% same (longer life) 732-9607-1-ND

1 CR9 MR756 6A; 600V 10A; 1KV 10A07-TDICT-ND

1 CR15 STTA12-060 600V; 12A; 55ns 600V, 16A, 50ns FES16JT-E3/45GI-ND

2 CR91,CR92 1N4006G-T 800V; 1A same 1N4006GDICT-ND

1 CR96 1N4936-E3/54 400V; 1A; 200ns same 1N4936-E3/54GICT-ND

1 CR97 MUR1100E 1A; 1KV; 100ns same MUR1100EGOS-ND

1 R1 1M; 1W; 1% 1M; 3W; 1% 749-2118-1-ND

1 R7 SL22 7R010 7 OHM 20% 10A 22MM same 570-1055-ND

2 R87,R90 ROX5SSJ27K 27K; 5%; 5W same A142834CT-ND

2 Q3,Q6 24N50 24A; 500V 30A; 600V IXFQ30N60X-ND

Notes:

1) The Martek designers get an ‘F’ in design for maintainability. Q3, Q6, and CR15 are mounted in the least

accessible way possible. Removing them is a pain. Getting them reinstalled with appropriate heat sink thermal

bonding and electrical isolation requires care and patience. Make sure you use non-conductive thermal paste.

Do not install C18 until Q3, Q6 and CR15 are installed and checked.

2) The PSU PCB is a .125” board with heavy copper plating. This means that anything connected to a ground plane

is a pain to desolder and solder. Solder wick won’t cut it; you will need a clean Hakko FR-300 (a 1 mm nozzle

worked best for me) or similar tool to remove components, and a high thermal mass temperature-controlled

soldering iron to install them. Especially for components that have trace connections on the top layer, make sure

that the solder fully bonds pad and pin.

3) Some of the components in the table above are larger than the components that they replace. They all fit, but a

bit of fussing (and in some cases, a bit of spacing above the board) is required.

4) If you use the 800V, 35A bridge, you will need a longer screw to reattach its heat sink (it’s a little thicker than the

lower amperage variants). A 6/32 x 1” stainless screw works well.

Final Testing:

The board came up without issue. All voltages, as measured on J1 and J2, were nominal (after reinstallation; the PSU

requires load). Here is a picture of the repaired, slightly better than before, PSU.

Here is a picture of the working TDS7404. Woot!