Drop out of MOSFET driver chip?

[Shocker]

Hi,

I have a MAX17601 (Datasheet) MOSFET driver chip. As far as i'm aware, it doesn't state a recommended maximum frequency or anything of a similar manner.

I've found that when i drive the chip with an input frequency of 40MHz, it works pretty well. The output amplitude stays pretty constant. When i push the frequency up, i find that the output of the chips starts to drop out at mid 50MHz. When i speak of frequency, i am referring to an on and off period of equal lengths.

This graph shows an example of the output of the driver without any MOSFETs or other loads connected to the outputs. Green = 40MHz, Blue = 60MHz.



As you can see, the driver is only being driven from some point between 0 and 100ns(1x10^-7). At 60MHz, i am only able to get 3 or 4 cycles before the output drops out. At first i thought there might be a PSU(a good benchtop PSU) issue, where the driver was depleting the local capacitors and the PSU wasn't kicking in quick enough to continue delivering energy to the circuit. I soldered a 33uF capacitor to the supply lead of the driver chip, to attempt to resolve this but it made very little difference.

I'm wondering if there are any other tricks or tips to get the driver chip working at the higher frequencies? Or just generally overlooking something?



Thanks

[Dago]

I'm not sure what you are trying to do but I'm fairly certain no one envisioned these MOSFET drivers to be driven at these kinds of frequencies. Switch mode power supplies usually top off at a few MHz.

My guess is that due to inadequate decoupling you are possibly latching the devices due to ringing in the supply. You need something like a small ceramic capacitor right next to the chip with good grounding and then something larger such as a big ceramic cap next to it. Or then the chips just won't work at these kinds of frequencies.

[Ian.M]

What do you expect when you are driving the input at an edge rate exceeding that of the propagation delay through the chip?  If there are any tweaks to speed up the edges of the internal level shifting, its not unlikely that some settling time is needed before the next edge can be handled properly.

[T3sl4co1l]

Yeah, that it does anything at all is fairly impressive; the power dissipation just due to the charging of internal capacitances will be quite significant.  And that probably causes unexpected voltage drops across internal nodes, causing skew and delay.  If nothing else, internal nodes won't be fully transitioning from '1' to '0', which tends to have the observed effect as well.

There are proper RF drivers out there, like by IXYS RF.  Mind I haven't seen any in stock anywhere...

Tim

[Shocker]

Does anyone recommend a MOSFET driver chip that is capable of driving a MOSFET at +50-60MHz? Or suggest a circuit that might be capable of such a frequency?

I have monitored the power line to the chip with a 1GHz oscilloscope and whilst i can obviously still small voltage ripples, it doesn't ripple more than a few hundred millivolts. Assume they'll be large ripples at even higher frequencies.

I have a relatively small ceramic capacitor on the PCB but it connects to the driver using a via. When i saw a voltage ripple on the supply line, i added the 33uf electrolytic on to the lead of the driver. Not ideal, i know, but it's kind of the best i could do with the current PCB.

I thought that a propagation delay was just the time taken for the input signal to make an effect to the output. So you could bang the input on and off as quickly you wished and all signals would be delayed by the propagation delay, which in my case is 12ns.

So i guess there is nothing i can do with my MOSFET driver to solve my issue?

I checked out IXYS RF drivers. I notice that their IXZ631DF18N50 datasheet states a maximum of 27MHz. Why is that when it appears that it could go much faster? - looking at rise and fall times.

I had checked out RF drivers before, but i thought they were limited in voltage. Looking at the IXYS one, i now realise that some RF drivers allow for a small voltage input signal and large output/drain-source voltages. Are there any other manufacturers of RF driver/RF driver and MOSFET module out there? I want over 50MHz switching at +20V outputs minimum. Ideally, i want higher output voltage and at up to 100MHz.

[Dago]

What are you trying to achieve? What type of a MOSFET are you driving? Just as an example driving an IRF510 (gate charge of 8.3 nC) at 50 MHz would require 830 A of drive current... Just to toggle the gate at that type of a frequency.

[Shocker]

I'm attempting to create a transducer driver, where i'm hitting it hard on and off. I'm doing it with a MOSFET from Diodes Incorp ZXMN10A07F. I'm only exciting the transducer for up to 200ns at a time. I think i calculated that the instantaneous current would be 8-9A with the MOSFET that i'm using.

[Shocker]

I'm only exciting the transducer for up to 200ns at a time.

For 200ns on time, assuming you are doing 50% duty thats 2.5MHz not 25MHz. Thats much more reasonable.

No. +50MHz for up to 200ns at a time. Eg. 10 cycles at 50Mhz = 200ns.

[Shocker]

20MHz is old news - it's been able to be obtained for at least 5 years now. There are different methods to drive a transducer, which i'm looking to replicate one of the more advanced methods at a higher frequency - hopefully around 50MHz.


Indeed. I agree with the MOSFET driver, which is why i asked about how to fix the issue. Or how i should go about designing a driver for such a frequency.

[T3sl4co1l]

Ah, so pulsed RF.  Do it the same way it's always been done -- big transistors, matching networks and tuning!

If it's not unreasonable to have a different envelope (like a ringdown), and poorer frequency stability or more harmonics, you might use a pulse forming network and compressor.  The RF equivalent of a Tesla coil, though usually phrased in terms of sharpening circuits (anything nonlinear, ranging from diodes and transistors, to varactors, saturable reactors and so on), or frequency mixing (varactor multipliers, varactor or step recovery comb generators..).

Tim

[Shocker]

Ah, so pulsed RF.  Do it the same way it's always been done -- big transistors, matching networks and tuning!

If it's not unreasonable to have a different envelope (like a ringdown), and poorer frequency stability or more harmonics, you might use a pulse forming network and compressor.  The RF equivalent of a Tesla coil, though usually phrased in terms of sharpening circuits (anything nonlinear, ranging from diodes and transistors, to varactors, saturable reactors and so on), or frequency mixing (varactor multipliers, varactor or step recovery comb generators..).

Tim

You sound familiar with the RF world. Do you have any suggestions/recommendations about what they use in the RF world to convert small signal (1-3V) to large signal (+30V). I know that 30V isn't particularly large voltage, but i'm trying to leave it as broad as possible.

[dmills]

The RF world does not really tend to think in terms of voltage, but in terms of power into a known load impedance (It amounts to the same thing of course, but for various reasons (complex impedances and transmission line effects, power is easier to do sums on).

What is the impedance of the transducer you are hoping to drive, and is it resistive or is it something like a piezo composite with a large fixed capacitance, what is the transducer loaded Q (This will determine the available bandwidth from the transducer and thus an upper limit on required bandwidth)? 

What envelope shape do you need?

Voltage transfomations up there are usually LC networks, sometimes transformers, sometimes carefully measured transmission line segments, power amplification is usually (These days) LDMOS running with some amount of standing bias, these are not (usually) hard switched.

Regards, Dan.

[T3sl4co1l]

If it's at the same impedance, then going from 1V to 30V simply requires 29.5dB gain.  Probably two stages, maybe three.  That could be done with a bandpass or wideband characteristic; it doesn't really matter, as transistors are more than fast enough to do that.

30V RMS into 50 ohms is 18W, so you'll need a modest sized power output transistor to get there, and a ~couple W 'predriver' for it.

If the load impedance is lower, you need that much more power.  If your signal source is, for example, LVCMOS or thereabouts, it'll be expecting a load over 100 ohms, and therefore will need several more gain stages to reach the output.  That's fair enough -- it's small signal stuff.

On the upside, high power transistors are almost all at modest voltages (~28V DC, which almost always means >56V peak by the way), so the collector/drain impedances tend to be very low indeed, requiring matching networks to reach useful transmission line impedances.  Often using planar transmission line segments to do the matching.  Which are necessarily quite narrowband, hence why you rarely see example application circuits claimed for wideband use.  Anyway, if your load is already a low impedance, you can just extend that super fat transmission line up to the load.  But the load itself better have a wide footprint, otherwise its few nH will kill the frequency response (and probably nuke the driver).


And going back to MOSFETs, don't forget that the drain needs to be charged and discharged as well.  That consumes as much peak reactive power (dissipated as loss, during turn-on) as the transistor's SOA can possibly muster, leaving none for the load!  If it can't discharge itself that fast, you absolutely cannot do any better at the load!

And if you're already driving that much power into the gate just to convince it to move, you can't possibly hope to get any more out of the drain because it simply doesn't have any gain at that speed!  Yes, don't be afraid to apply concepts such as gain and loss at frequency, or gain bandwidth product, even when the device is being operated with very large signals.  Fundamental limits are fundamental, whether they're being observed by small signal parameters and amplifier efficiency, or full signal swing with switching speed and efficiency.

Tim