Yes, I have used both Analog devices and TI Digital Isolators to perform the level shifting task in floating gate drives. They work very well the only problem is there is the need for a 5V regulator, as these are 5V or less logic. The TI ISO722 is one of my favorite devices to use, it has high transient immunity, very high speed, low pulse width distortion and very good isolation rating.
Hi David,
Have a look at their versatile audio driver chip, only thing of concern is that their method of operation involves transferring RF energy at 16Mhz to achieve isolation and signal transfer across different potentials ,that might act as antenna and plague the EMI performance in Mhz region.
https://www.silabs.com/products/audiovideo/class-d-audio-drivers/Pages/Si824x.aspx
Kanwar
Have a look at their versatile audio driver chip, only thing of concern is that their method of operation involves transferring RF energy at 16Mhz to achieve isolation and signal transfer across different potentials ,that might act as antenna and plague the EMI performance in Mhz region.
https://www.silabs.com/products/audiovideo/class-d-audio-drivers/Pages/Si824x.aspx
Kanwar
Hi, Photonicube,
Which one is better, fast optocoupler or digital isolator?
Kanwar mentioned about their ClassD high/low side driver. Si824x seems to use these digital isolators for both high side and low side signal transfer.
Hi, Kanwar,
Interesting driver. I haven't read very carefully, but the adjustable dead time is very interesting. IRF counterpart have only some selectable values.
Which one is better, fast optocoupler or digital isolator?
Kanwar mentioned about their ClassD high/low side driver. Si824x seems to use these digital isolators for both high side and low side signal transfer.
Hi, Kanwar,
Interesting driver. I haven't read very carefully, but the adjustable dead time is very interesting. IRF counterpart have only some selectable values.
Hi Lumanauw,
I tend to favor the digital isolators over optocouplers, in most cases, for a couple of reasons. The first being the input impedance of the digital isolators is very high, making high speed driving and terminating of this input much easier. Second being optocouplers current gain reduces over time, meaning you have to pump more current into them to get the same output. The third reason is, alot of my designs must operate in a 105C ambient enviroment and I have not found any decently fast optocouplers that can handle this. Optocouplers do have their place, they are very good in highly noisy enviroments, Analog devices chips can not be placed near strong varying magnetic fields like transformers or inductors, optos can. The SIlabs parts are suseptable to high voltage slew rates so, I.E. radiated EMI, as stated by Kanwar about. TI's capacitivly coupled technology is the closest to optocouplers while hitting 150Mbps+ speeds, hitting the 125C junction temperature, having high transient voltage immunity and achieving very high isolation voltage rating. The main problem I have with the TI part is the price $4 or $5 per chip or $13 for MIL-Spec version, gets expensive if you have a board that needs 10 or 12 of these. The really nice part of all this is, the single channel versions of Analog devices, TI, SIlabs, NVE, AVAGO, etc.... isolation parts have the same SO8 pinout, this means you can layout a board and findout first hand which one you prefer, that is what I ended up doing. You may find that one works better in your application than another or if they all work equally well in your circuit than pick the cheapest
I tend to favor the digital isolators over optocouplers, in most cases, for a couple of reasons. The first being the input impedance of the digital isolators is very high, making high speed driving and terminating of this input much easier. Second being optocouplers current gain reduces over time, meaning you have to pump more current into them to get the same output. The third reason is, alot of my designs must operate in a 105C ambient enviroment and I have not found any decently fast optocouplers that can handle this. Optocouplers do have their place, they are very good in highly noisy enviroments, Analog devices chips can not be placed near strong varying magnetic fields like transformers or inductors, optos can. The SIlabs parts are suseptable to high voltage slew rates so, I.E. radiated EMI, as stated by Kanwar about. TI's capacitivly coupled technology is the closest to optocouplers while hitting 150Mbps+ speeds, hitting the 125C junction temperature, having high transient voltage immunity and achieving very high isolation voltage rating. The main problem I have with the TI part is the price $4 or $5 per chip or $13 for MIL-Spec version, gets expensive if you have a board that needs 10 or 12 of these. The really nice part of all this is, the single channel versions of Analog devices, TI, SIlabs, NVE, AVAGO, etc.... isolation parts have the same SO8 pinout, this means you can layout a board and findout first hand which one you prefer, that is what I ended up doing. You may find that one works better in your application than another or if they all work equally well in your circuit than pick the cheapest
In a compact design, this level shifter will be placed near of classD output inductor. Will this be a problem for a Digital Isolator?
SiliconLabs Si8410 is using RF modulator/demodulator for transfering data. TI's ISO722 datasheet stated "a periodic update pulse is sent accross the barrier to ensure the proper DC level of the output"
Iis ISO722 also use somekind of RF communication between it's input and output, like Si8410?
Propagation delay of Digital Isolator is fast, Si8410 below 10nS, ISO722=17nS, while 6N137=50nS.
SiliconLabs Si8410 is using RF modulator/demodulator for transfering data. TI's ISO722 datasheet stated "a periodic update pulse is sent accross the barrier to ensure the proper DC level of the output"
Iis ISO722 also use somekind of RF communication between it's input and output, like Si8410?
Propagation delay of Digital Isolator is fast, Si8410 below 10nS, ISO722=17nS, while 6N137=50nS.
So far I experimented with the digital isolators from Agilent only.
HCPL9030 showed really nice pulse handling.
Refering to posting #89 & #90
http://www.diyaudio.com/forums/class-d/101774-1kw-gen2-9.html
But finally I did not dare to use them in a real amp, because their specified dv/dt is 15kV/us and calculations by hand and simulation as well showed that in my next amp the resulting dv/dt can reach values up to 30kV/us....
I think the max dv/dt immunity of such isolators is one of the key parameters besides propagation delay and output rise times. The silicon lab series Si824xx is looking pretty promising for me. Thanks for pointing towards these devices.
HCPL9030 showed really nice pulse handling.
Refering to posting #89 & #90
http://www.diyaudio.com/forums/class-d/101774-1kw-gen2-9.html
But finally I did not dare to use them in a real amp, because their specified dv/dt is 15kV/us and calculations by hand and simulation as well showed that in my next amp the resulting dv/dt can reach values up to 30kV/us....
I think the max dv/dt immunity of such isolators is one of the key parameters besides propagation delay and output rise times. The silicon lab series Si824xx is looking pretty promising for me. Thanks for pointing towards these devices.
50kV/us?
Only if you have good luck.
For a reliable design I tend to anticipate the typical values for uncritical functions only. But if things are coming to the decision whether one Fet is ON or may be both...then I tend to work with the guaranteed values.
Means 25kV/us for the ISO722, which is still very good.
The charming thing with the silicon lab devices is the high integration for both half bridge switches including drivers and dead time adjustment.
This makes me thinking... I already have the IR20957 on the desk and first tests showed that I probably can live their performance, but the Si824xx are temptating...
But probably I will stay with the IR - even the overcurrent protection seems to work acceptable, so that's an advantage.
Only if you have good luck.
For a reliable design I tend to anticipate the typical values for uncritical functions only. But if things are coming to the decision whether one Fet is ON or may be both...then I tend to work with the guaranteed values.
Means 25kV/us for the ISO722, which is still very good.
The charming thing with the silicon lab devices is the high integration for both half bridge switches including drivers and dead time adjustment.
This makes me thinking... I already have the IR20957 on the desk and first tests showed that I probably can live their performance, but the Si824xx are temptating...
But probably I will stay with the IR - even the overcurrent protection seems to work acceptable, so that's an advantage.
So far this did not see any critical impact from that, but I was not doing a full amp and also no EMI measurements.
I stoped, when seeing that I would run into issues with dv/dt.
I am doing a LTSPICE sim of a SIC amp using UCC21521C and the IC is doing quite well. (Main problem is the extremely slow simulation engine, no matter if I use my dual 8 core Xeon or my i7-8700K. Have heard that there is a new 56 core Xeon out now of which several can be paired on a mainboard. But $ 70K+ for a computer would be a bit steep....)
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.