Truepath: Soldering parts directly to MOSFETs

[audiophool]

From reading the TA3020 datasheet and the corresponding reference board datasheet, it would appear that certain capacitors and diodes should be soldered directly to MOSFET leads rather than mounted on the PCB in the normal manner. Apparently not doing so reduces the reliability and perhaps the sound quality of the reference board.

As I understand it, this situation applies identically to the Truepath because the Truepath schematic is component-for-component identical to the schematic of the reference board (apart from the inclusion of the mute auto recovery circuit in the Truepath which is not included in the reference board.)

The TA3020 reference board has the following notes on the schematic (refer to page 15 of www.tripath.com/downloads/RB-TA3020.pdf)

***********
D106, D107, D206, D207 ARE MUR120 DIODES SOLDERED DIRECTLY TO FET LEADS ON BOTTOM OF BOARD
C113 AND C213 ARE SOLDERED DIRECTLY TO FET LEADS ON BOTTOM OF BOARD
***********
The TA3020 datasheet (pages 7-8) refers to D106, D107, D206 and D207 as DO and says:
"These diodes absorb any high frequency over/ under shoots caused by the output inductor LO during high output current conditions. An ultra fast recovery diode that can sustain the entire VPP-VNN voltage should be used."

Likewise, the TA3020 datasheet refers to C113 and C213 as CHBR and says:
"These components must be located as close to the output MOSFETs as possible to minimize output ringing which causes power supply overshoot. By reducing overshoot, these capacitors maximize both the TA3020 and output MOSFET reliability."

Although the difference in inductance between PCB mounted components and FET mounted components would seem to be negligible, it is clear that Tripath considers this marginal minimization of the inductance of these capacitors and diodes to be very important, otherwise, why would they include this matter in their documentation knowing that it could result in significant increase in OEM product design and manufacture? I guess it could make the difference between amps used at high power lasting 2 years and 5 years before failing - something like that. I expect that this issue becomes even more important in cases where people have left MOSFET leads longer than necessary in order to accommodate particular heatsink arrangements. For example, I have seen photos where the leads are bent 90 degrees and the MOSFETs are mounted horizontally on the heatsinks, in which case the leads are perhaps double the length they would be if the MOSFETs were mounted vertically.

Im wondering has anyone followed Tripaths advice? Also I would be interested in whether there is any audible difference if the components are soldered directly to the MOSFET leads.

Tim

[V-bro]

Truepath has a four layer board and much shorter leads to the mosfets than the reference board. Therefore these components can be mounted on the PCB without problems.

------X-max. can be several meters on any driver, too bad it can be done only once...------

[Scratchy]

I am one of those that are bending the Mosfet leads 90° and have often wondered the same thing for the AMP2 reference design boards. I have already soldered the diodes but was considering soldering the caps directly to the Mosfet leads under the PCB.

Comments ???

[V-bro]

AMP1, Mosfet leads??

[audiophool]

Scratchy,

I have always been a bit puzzled by Tripath's requirements as I laid out in my initial post on this thread. I think I can shed some more light on them now, but first I want to dispel the following piece of misinformation.

Truepath has a four layer board and much shorter leads to the mosfets than the reference board. Therefore these components can be mounted on the PCB without problems. ...

Premise 1. Truepath has a four layer board: the layers do not significantly affect the impedance of the gate circuit and hence this premise is not relevant.

Premise 2. Truepath has much shorter leads to the MOSFETS (presumably this means gate the connection to the gate): Take a look at the Evaluation board PCB drawings in the appendices of the datasheet and you will see that this is not the case, especially when taking into account the length of RG and the return path.

Conclusion. Therefore these components can be mounted on the PCB without problems: Premise 1 is irrelevant. Premise 2 is incorrect. Hence this conclusion is void.

[V-bro]

Premise 1. The four layer board certainly has lower trace inductance hence this premise is perhaps more relevant as it seems.

[Scratchy]

Here is a quote from the RB-TK2350 PDF:

The output MOSFETs (M1, M2, M3, and M4) provide the switching function required of a Class-T design.
They are driven directly by the TP2350B through the gate resistors. The devices used on the reference board
are ST STW34NB20 MOSFETs. The TK2350 data sheet contains information on output FET selection as well
as Tripath application notes “FETs – Selection and Efficiency” and “Designing with Switching Amplifiers for
Performance and Reliability”.
The bypass capacitors C105/C205 and C113/C213 are critical to the reduction of ringing on the output
MOSFETs. These parts are placed as closely as possible to the leads of the MOSFETs, and the leads of the
capacitors themselves are as short as practical. Their values will not change with different output MOSFETs.
The output diodes D106/D107/D206/D207 are also critical to the reduction of ringing on the outputs of the
FETs. They shunt the inductive energy if the output exceeds VPP or goes below VNN. The proper connection
of these diodes are “drain to drain” and “source to source” as shown in the schematic diagrams.

Strange thing is, those Diodes and Caps do not exist on diagram 8 nor in the reference design which is exactly what the AMP2 is based on. Perhaps those references are for the TP reference design and they did not change them?

In any case, due to space constraints, the leads on my mosfets are very long (longer than I would like).
I'll snap a pic once I get it mocked up.

[V-bro]

"They shunt the inductive energy..."

I knew I read that somewhere, thanks Scratchy.

[V-bro]

It is quite puzzling matter indeed, these diodes seem to be present only named differently? In the TA3020 schematics there are no diodes mounted over the fets whatsoever....

[audiophool]

The bypass capacitors C105/C205 and C113/C213 are critical to the reduction of ringing on the output
MOSFETs. These parts are placed as closely as possible to the leads of the MOSFETs, and the leads of the
capacitors themselves are as short as practical. Their values will not change with different output MOSFETs.
The output diodes D106/D107/D206/D207 are also critical to the reduction of ringing on the outputs of the
FETs. They shunt the inductive energy if the output exceeds VPP or goes below VNN. The proper connection
of these diodes are “drain to drain” and “source to source” as shown in the schematic diagrams.

Scratchy, The latest version of RB-TK2350 is Rev2.2, and this is what it says:

The bypass capacitors C3/C4 and C6/C7 are critical to the reduction of ringing on the output MOSFETs.
These parts are placed as closely as possible to the leads of the MOSFETs, and the leads of the
capacitors themselves are as short as practical. Their values will not change with different output
MOSFETs.
The output diodes D21/D23 and D22/D24 are also critical to the reduction of ringing on the outputs of the
FETs. They shunt the inductive energy if the output exceeds VPP or goes below VNN. The proper
connection of these diodes are “drain to drain” and “source to source” as shown in the schematic
diagrams.