Dear Michael,

Technical view first:

Concidering power modules for applications >>10kW, I think that the future of DCB is simply DCB.

The omnipresent trendency of increasing both, power density and reliability, demands a robust carrier for chips. Additionally, new interconnection technologies to eliminate soft soldering of chips need a base material that is resistant to heat >>300°C and high pressure.

Furthermore, thin material do not help in spreading heat and thus will worsen the situation for thermal management issues.

For power devices up to 10kW of throughput the world is different as DCB is not used here for all devices. There are good designs out there featuring lead frames and I don't see pressure here to change this.

If however, thin film technology would remain compatible to today's - and future - demands while reducing production cost, a closer look would be needed.

Finally it is all about one single thing: Money spent.

Historical view:

The 62mm power module was developed in the 1990s and still is a high-runner in industrial applications.

You can consider this a hint that a well-established design will remain out there for at least 30 years. As there is no thin-film based power device today but several new devices currently under development around the world, I conclude that DCBs will remain the dominant technology for the decade to come.

Dear Martin,

thank you for your comment. I think, you mixed up thick film with thin film - which makes a huge difference. You brought the point of temperature stability - thick films are produced by sintering screen-printed pastes at about 950 °C and are not containing organic material after the firing process. This makes them really good and sustainable insulators. You further adressed current carrying current and heat - this depends significantly on the ceramic substrate used. DCBs of AlN behave much better than the more widely used Al2O3 substrates. But the same is true for thick-film substrates, where the substrate also mainly defines the thermal properties. The requirements regarding current carrying capability need to be adressed by thick enough copper layers - which is possible with more than one layer of copper conductor paste deposited on the substrate. The reasonable maximum film thickness is about 60 µm - and you could build up to 300 µm where needed. In other words you could combine areas of high current carrying capability and areas where fine structures allow for complex circuits with small pitch.

I understand well that all is about value for money - but that is exactly my point: Why not taking more out of the module concept?

There are some few modules basing on thick-film substrates offered by Vincotech. Don't you think that wide-bandgap semiconductors will change the rules of the game significantly?

Hi Michael,

Nope - I wasn't wrong about thick/thin film. I simply stated "something thin" and referred to anything thinner than DCB ;-)

Kidding - sorry...

However - about your numbers:

Insulation of 60µm at high voltages of UDC>600V?

How about the field stress here?

Dear Martin,

right, the insulation toughness is a main issue in case you want to go for high-voltage multi-layer circuits within the module. We are currently investigating that and work on ways to transfer thick-film technology to modules of the 800 V class. What's feasible for sure is low-voltage insulation of 60 V, which greatly helps to lower gate circuit inductance and shrinks the space requirements of related circuitry. As I stated, thick-film goes up to 300 µm - it is likely that reliable insulation using multiple 60 µm ISO-layers is possible. And further, there is a huge difference whether one wants to insulate the half-bridge output to ground or just the DC-link. That is because DC with a small low-frequency content is much easier to reliably insulate than HF high voltage ... .

What else would you put under question?  

Hi Michael,

What else..... let me think...

How about mass production compatibility?

Solid DCBs are easily handled, robust in assembly lines, fall into place without bending

Ok - if you consider your film to be the layer between two copper layers of 300µm thickness, the result would be similar in stiffness, right?

What about adhesive capabilities regarding thermal swing, namely thermal cycling.

How about CTE-mismatch between your film and copper or ceramics?

And finally - how resistive is the film in regards of partial discharge?

Seems to me, there is a lot to consider. As a consequence, I see several hundred man month of testing and qualifying coming up :-)

Dear Martin,

right, there will be a lot work to do until everything is "industry-ready". However, be aware that traditional hybrid circuits based on silver conductor films are very well-known for >30 years and are still the #1-source for high-reliability circuits and systems.

Basically, copper thick-film for modules is nothing else than another conductor (copper), new compatible insulators and much larger thicknesses. Nobody wanted to build up silver up to 300 µm exept for producing high-priced jewellery.

Mass production:

Nowadays, industry is very relying on less than a hand full of DCB manufacturers with differing quality standards and pricing. As far as I am aware, one could wait for months to receive a new design.

Differently, thick-films base on standard ceramic sheets (for instance 0.625 mm thickness) and are produced with additive processes - no etching required. The needed screens with the intended patterns can be produced in house within hours, several substrates can be printed in one run. It seems that this technology is really suitable for manufacturers who really want to build customized modules that exactly fit to the (dynamically changing) needs of their customers. Main cost drivers are the pastes (economy of scale) and the high-temperature firing process (economy of scale / ovens with heat recovery).

Stiffness mainly depends on the properties of the used substrate. If also the back side of the substrate is screen-printed, one could achieve the same mechanical sandwich structure as the DCB. As far as our work shows, current carrying capability and thermal cycling capability can lay close to that of DCBs. Since printed copper and insulators need to withstand multiple cycles 25 - 950 °C, the CTE mismatch is optimized and should - with proper design - not impose reliability problems. Moreover, since different film thicknesses are possible, tension peaks can be prevented ... .

I wait for more ... comments, critics ... everything is welcome!   

Hi Michael,

there was the Cooling Days Congress in Würzburg/Germany this week and I talked to someone about thick-film, screen printed, sintered, multi-thickness, DCB-like structures and getting my hands on a sample in the exhibition made me think...

Technology-wise it looks promising and what I find most intriguing is the option to combine 30µm copper in 250µm design-rules with 300µm copper for high-power components.

Hello Martin,

actually that is exactly what inspires also my colleagues and me. We are currently working on processes and their fine-tuning in order to get best thick-film printig results. We already combine high current areas with 100 - 200 µm with fine-print areas small pitch. "Multi-thickness" is an adequate word for it - thank you for inventing it. 

Also if you think on high-frequency operation of power modules and the power lost in the driver and in the gate resistors, it seems to be wise to implement the driver on the module substrate. The heat transfer achievable with the ordinary PCB is quite restricted, isn't it? 

The copper thick-film (CTF) technology definitely has something in common!

Martin, may I also refer to our publication at the ESTC 2014 "Copper Thick Film Substrates for Power Electronic Applications". Further publications at the PCIM 2015 and EPE 2015 are on schedule.

Hi Michael,

1) Looking forward to seeing you in Nuremberg and Geneva :-)

2) There might be a drawback with printed/sintered materials regarding material density. I recently got the info, density and thus conductivity are down to 80% of the copper used in DCB-materials.

This will increase resistive losses, especially in high-power applications. It will also change current carrying capability of the layer and thus will lead to wider structures which may turn out to be detrimental for the layout of high-power devices.

I'd be curious to what will develop in the next years to come. 

Dear Martin,

this 80 % density of copper thick-film compared to solid copper is not far from reality. As you pointed out, this may indeed limit the current carrying capability when it comes to very high power applications. However, very high power applications won't run at 1 MHz switching frequency in the near future.

Copper thick-film may close the gap between the molded IPEMs and the classic DCB-based medium to high power modules. In the range of tens to hundreds kW we have that chance to combine both: sufficient current carrying capability and  a high level of integration. This will push possible switching speed and thus either reduce losses or/ and reduce system size.

In many modules, the 300 µm copper thickness of standard-DCBs may not be required at any place of the substrate. Copper thick-film may ensure suitable thick layers where needed and simultaneously provide multi-layer structures to house gate-drivers, ceramic DC-link caps and so on ... .

I appreciate your interest a lot and look forward to meet you at the upcomming events!