Just received my copy of Bob's "Designing Audio Circuits and Systems" book and wanted to send a huge THANK YOU to @Bob Cordell .
I'm sorry to be so late in responding to this post. For some reason it has been spotty in my receiving notifications about this thread by email.
I've used the IRFP240 and IRFP9240 numerous times, and they have performed very well. They have very high current capability and have high transconductance. I used vertical MOSFETs like these back in the '80s when I did my MOSFET power amplifier with error correction. That JAES paper is available on my web site. These devices are very fast, and care needs to be exercised in the layout, bypassing and circuit design to avoid parasitic oscillations. Those issues are touched upon in that paper. They are not trenchfets, which are designed for extremely low Rds_on in SMPS applications and do have a bit of a secondary breakdown issue when used in a linear design.
I chose the vertical MOSFETs over the laterals because the verticals are capable of higher performance, with greater transconductance and lower Rds_on. That issue is also touched upon in my JAES paper. The laterals are also a bit slower because their construction causes them to have higher effective series gate resistance. On the other hand, they do tend to have lower drain-gate capacitance than the verticals, especially when Vdg is smaller.
Nevertheless, Rick Savas and I used them in designing the redesigned DH-220C, since we were using a DH220 as the infrastructure for the design, including the original lateral MOSFETs that it employed. We did achieve excellent performance with that design, so there is certainly no problem with using Excicon laterals in a design.
Cheers,
Bob
You are correct about the P channels not being an exact complement to the N channel vertical MOSFETs. This is largely due to the difference in mobilities between P and N device characteristics. As a result, transconductance is a bit different at a given drain current. I often use small source resistors in series with the sources, but not for bias stability or reduction of current hogging etc. as for BJTs. These resistors are in the tenths of ohm range if I recall. I discovered that if I deliberately use different values of source resistors for the N and P channel devices I can somewhat mitigate the effects of these differences, which are mainly in the class AB crossover region. I discovered this in simulation by doing wingspread curves.
BTW, there is no significan optimum idle current for MOSFETs to minimize crossover distortion as there is for BJTs, e.g. the Oliver Criterion where in principle you bias the output stage for 26 mV across each emitter resistor (the real-world optimum is often closer to 20 mV). For MOSFETs, the more bias current the better, as long as thermal conditions are OK. Instead of gm doubling with higher bias current as you get with BJTs, you generally get gm droop in the crossover region due to the lower transconductance of the MOSFETs. Overall, however, the crossover in a MOSFET design is smoother.
Cheers,
Bob
You are very welcome. Thank you for buying my new book and for your kind words.
Cheers,
Bob
Hi Bob,
I received this latest edition a short while ago. It is next on my reading list.
Vertical Mosfets do have a gate charge phenonium, I've found that driving them with a higher current driver stage is best. As you know, most early commercial designs drove them with vacuum tubes or TO-92 devices at low currents. Not a good plan.
I received this latest edition a short while ago. It is next on my reading list.
Vertical Mosfets do have a gate charge phenonium, I've found that driving them with a higher current driver stage is best. As you know, most early commercial designs drove them with vacuum tubes or TO-92 devices at low currents. Not a good plan.
Yes, you are correct. As a source follower configuration in an output stage, the gate-source capacitance is not a big problem because it is boostrapped by the output. However, the gate-drain capacitance can be an issue if the gate is driven from a high-impedance source, as you correctly point out. That capacitance is nonlinear at smaller gate-drain reverse voltages, as when the voltage swing is high and comes toward the rail. In some arrangements, that increased gate-drain capacitance can cause some local instability as well.
I always drive the MOSFETs, be they vertical or lateral, with an emitter follower, with just a gate-stopper resistor in between. If one wants to go to a lower value of gate stopper resistor to get the most speed and least output phase lag, perhaps as low as 50 ohms, then one must use a gate Zobel network to ground on each gate. In cases with paralleled devices, this needs to be done for each gate. The use of the gate Zobel network is discussed in my JAES paper describing the MOSFET power amplifier with error correction which can be found on my website. Note that even the Hafler lateral MOSFET designs include a gate stopper resistor of at least 220 ohms.
Stability problems with vertical MOSFETs are usually caused by the unintentional formation of a Hartley or Colpitts oscillator involving the parasitic inductances (internal and external to the MOSFET) and the MOSFET's capacitances. The gate Zobel network damps these potential resonant circuits. There is one more related issue. MOSFETs have a thin gate oxide that can only withstand 15 or 20 volts If this is punctured, the device gets destroyed and a drain-source short results. Zeners are often placed gate to source to prevent such overvoltage, or drive circuits that inherently cannot create such gate-source voltages are usually used.
HOWEVER, if the device breaks into oscillation due to an inadertant oscillator configuration involving inductance, the resonance can result in higher voltage swings at the internal gate than at the gate pin, and the external gate-source limiting may not be enough to prevent over-voltage at the gate itself internally. As we all know, a series-resonant circuit can amplify the voltage at resonance.
Cheers,
Bob
I always drive the MOSFETs, be they vertical or lateral, with an emitter follower, with just a gate-stopper resistor in between. If one wants to go to a lower value of gate stopper resistor to get the most speed and least output phase lag, perhaps as low as 50 ohms, then one must use a gate Zobel network to ground on each gate. In cases with paralleled devices, this needs to be done for each gate. The use of the gate Zobel network is discussed in my JAES paper describing the MOSFET power amplifier with error correction which can be found on my website. Note that even the Hafler lateral MOSFET designs include a gate stopper resistor of at least 220 ohms.
Stability problems with vertical MOSFETs are usually caused by the unintentional formation of a Hartley or Colpitts oscillator involving the parasitic inductances (internal and external to the MOSFET) and the MOSFET's capacitances. The gate Zobel network damps these potential resonant circuits. There is one more related issue. MOSFETs have a thin gate oxide that can only withstand 15 or 20 volts If this is punctured, the device gets destroyed and a drain-source short results. Zeners are often placed gate to source to prevent such overvoltage, or drive circuits that inherently cannot create such gate-source voltages are usually used.
HOWEVER, if the device breaks into oscillation due to an inadertant oscillator configuration involving inductance, the resonance can result in higher voltage swings at the internal gate than at the gate pin, and the external gate-source limiting may not be enough to prevent over-voltage at the gate itself internally. As we all know, a series-resonant circuit can amplify the voltage at resonance.
Cheers,
Bob
I believe that this site only sends one notification per thread and then stops until you visit the site to get further updates on that thread. Thus, if you miss one notification in your email inbox, you probably won't get another, no matter how many posts are added to the thread. The solution is probably to periodically check the thread, and that should resume the notifications. This is all observed behavior - I hope the site explains the design somewhere, but I don't see a FAQ.
Thank you very much for describing this situation. I think this behavior of the site began after thay chaged the site management software quite awhile back. I am pretty sure it did not behave this way in the years before this software change. I had actually thought that activity on DIY audio had fallen dramatically. If you are correct about this issue, I would urge the folks running DIYaudio to fix this behavior that I consider to be a problem.
Cheers,
Bob
Cheers,
Bob
Hi Bob,
Yup. Still finding out about small differences in site software. Some things a LOT better, others are annoying. The previous software was heavily modified (we didn't look like other sites using the same software). Just migrating to a completely different platform was a job. It will only get better over time. Jason has done amazing work here.
All your points on Mosfet drive are correct. I include a G-S zener as well, even on regulator pass transistors. School of hard knocks (why did that short? lol!). Not very many Mosfet output sections use zobel networks, but a damping resistor is required every time. For the same reason as most tube amplifiers.
Yup. Still finding out about small differences in site software. Some things a LOT better, others are annoying. The previous software was heavily modified (we didn't look like other sites using the same software). Just migrating to a completely different platform was a job. It will only get better over time. Jason has done amazing work here.
All your points on Mosfet drive are correct. I include a G-S zener as well, even on regulator pass transistors. School of hard knocks (why did that short? lol!). Not very many Mosfet output sections use zobel networks, but a damping resistor is required every time. For the same reason as most tube amplifiers.
I believe that it is considered a feature. Anyone who participates in many threads would get many updates, and would otherwise need to turn off notifications for threads after they lose interest in those threads. This feature allows you to continue to get notifications for threads that you're actively reading, but no notifications if you never return to that thread.
Before this feature, I was getting a slew of emails about every thread I had ever commented in. After this feature, things are more manageable.
There are ways to improve this. Accounts could have the option to keep all notifications active, although that would generate a lot of email traffic and might be a burden on the site. Another option would be a way to subscribe to all updates on specific threads. Yet another option might be to enable notifications any time someone mentions your account or quotes you in a thread. Unless someone is getting quoted a lot, the latter feature might be quite effective.
Thanks for your follow-up. Everything you said makes sense. Your ideas for improving matters regarding this issue all make sense. Maybe there could even be an automated email warning before someone is cut loose from a thread.
Cheers,
Bob
Cheers,
Bob
There's actually a warning in your "final" notification. Here's an excerpt from the email that I received as a result of your reply:
Basically, they outline the contract right there: You gotta go read at least one new reply before you're notified of any more replies.
Granted, this paragraph is the middle of three paragraphs at the bottom of the notification email, so folks might not notice it. That's the problem with warnings, though - they often go unnoticed in today's Information Overload Atmosphere.
p.s. Thanks for the great books, past and present. I have a few in my library, and have read at least one, from cover to cover. Great stuff! Even a four-year BSEE degree doesn't cover all of the important details that your books cover, so they're greatly appreciated!
Up at the top of a diyaudio page are some menu items. One is to look at watched threads. Any thread you have replied to before should be in that list. You can choose to look at the list of watched threads, including any you have not fully viewed as of yet.
Thanks!
To your point about BSEE teaching, the late Marshall Leach was an EE prof at Geargia Tech. He wrote a good book covering audio electronics and acoustics. He put together a course and a lab based on his book, and it turned out to be an extremely popular course with the students. He wisely used audio as the hook to get students interested in EE in general. In the lab portion, the students completed many really good projects, one of which was an amplifier that he had designed and published.
Cheers,
Bob
That sounds great! Thanks for sharing that.
My alma mater is NCSU, where some of the student projects included designing and actually fabricating a 16-bit microprocessor chip! There were also hands-on labs where students could design and build digital circuits.
Alas, there was not very much in the analog domain. Before NCSU, I devoured Craig Anderton's Electronic Projects for Musicians. Nothing quite as powerful as an amplifier in that book, but plenty of small-signal audio projects there, complete with layouts for hand-etching.
Don't feel badly. The world of analogue electronics is one where you learn every day. You build on what you know and never stop.
There is no way you can put all this information even in a 10 year course. Plus some folks get it, some never will. The sad thing is that when you pass on, most of what you've learned hasn't been passed on. All you can do is pass on as much as you can.
There is no way you can put all this information even in a 10 year course. Plus some folks get it, some never will. The sad thing is that when you pass on, most of what you've learned hasn't been passed on. All you can do is pass on as much as you can.
I think hands-on student projects can make all the difference in the world. I learned an awful lot in the course labs at RPI where I went for my BSEE. Hands-on experience leads to more hands-on experience, and helps one acquire common sense and critical thinking ability. It gives context to the more theoretical material that one learns in the purely academic courses. A couple of weeks ago I was at Burning Amp and it was great to see so many people with soldering irons!
Cheers,
Bob
Hi Bob,
Do you mean Ryerson in Toronto? That's where I went.
As for hands-on experiments ... ABSOLUTELY! I loved the practical applications married with university level content. We built the chassis, PCB and everything for our projects back then.
Do you mean Ryerson in Toronto? That's where I went.
As for hands-on experiments ... ABSOLUTELY! I loved the practical applications married with university level content. We built the chassis, PCB and everything for our projects back then.
US university attendees generally use "RPI" as an abbreviation of Rensselaer Polytechnic Institute in Troy, New York. Here is its Wikipedia article. Bob Cordell's MOSFET amplifier article in the Audio Engineering Society Journal, says more.
_
_
