| RiskCord |
Hey guys.
I’m currently designing a 200 W Class-D amplifier, which in principle would switch at 500 KHz with a half-bridge output stage. As I decided not to use any dead-time compensation (current polarity sensing is very problematic) the challenge is to make the dead-time as short as possible (my project supervisor told me I should achieve a dead-time of less than 50 ns, including rise and fall times of the switches). Well, naively I thought the limitation on switching speed would be mainly imposed by the MOS but I found very fast switches such as IR’s Digital Audio MOSFETs (IRFI4020H). I realized instead that gate drivers are much more restrictive... Lets see, for instance IRS2011 specifies a rise time of 40 ns (max). HIP2100 are slightly faster but are for 100 V... not good enough. Anyway, any suggestions or ideas on this respect would be very welcome!
Thanks! |
|
|
| Eva |
Dead time is the period during which both switches are off and it's one of the major sources of distortion. However, very fast gate drive is not actually required to get a short dead time, it's actually a matter of matching propagation delays and gate charge/discharge rates so that each switch turns on just after the opposite one has turned off, sometimes with a slight amount of overlap.
With IR2010, 50ns DT is easily achievable, you could even aim at 10ns DT. Note that the amount of distortion produced by a given DT is proportional to the switching frequency and 500Khz is quite high, 250Khz would be more forgiving. Also, at such a high frequency you may experience too high dissipation in the gate driver IC, particularly if you use a SMD part. |
|
|
| lumanauw |
| Like EVA said, it's better to see the total final delay difference between high and low. This will include the comparator, driver and mosfet itself. Adjust from this final delay. In reality, the shape of the gate drive can make difference. If it is not perfect square (like a bit trapezoid due to rise time), it can use less dead time. |
|
|
| Rembrandt |
Would intersil's ISL6605 (or 6609) be an option?
(Also curious for myself, because I would like to design a class D with IRF7807Z, maybe combined with this driver). |
|
|
| RiskCord |
Thank you guys for your help. I'll try an IR2010 and play for a while with it to see how it performs.
The ISL6605 looks wonderful but it is intended for very low power applications, so its not useful to me, but it looks it goes just fine with the IRF7808Z |
|
|
| RiskCord |
I found this baby:
IRS20955pbf
PWM floating input, bidirectional current protection and programmable dead time.
The minimum dead time dispersion is highly improved when comparing to IRS20124 (+- 7 ns for the minimum dead time, 15 ns).
Has anyone tried it? Any comments? |
|
|
| Pafi |
| UCC27200. Really fast, powerful. For high power you can build a bridge. |
|
|
| RiskCord |
| Thanks Pafi. The UCC27200 won't make it sincec it has a maximum Vb of 120 V and my rail voltages are around +-65 V. The UcC project by Bender looks quite good indeed so I guess it wont hurt trying the IRS20955! |
|
|
|
