diyAudio Forums Archive > Top > Amplifiers > Class D Pages: [1] 2 3  Class D frequency? - Click HERE for Original Thread zenmasterbrian My own interest in class D is for subwoofer only in a multiamped system. So the class D amp would only have to go to 100hz, nothing more. So, it could sample as low as 200hz, but of course that is audible. So, maybe have a low sampling frequency and filter it out. Or, maybe has a higher sampling freq, but not as high as 40khz. So it is inaudible. What is the concensus on these topology and architecture issues. I've looked but if another thread deals with this, I'll go there.:rolleyes: SpittinLLama I'm not fully sure what you are asking but I don't know why you would want to sample the incoming signal so slowly. If you want to use that as a way to filter then you'd be better off using a simple op amp filter then sample at the higher frequency to get excellent resolution. If you're talking about the PWM oscillation frequency switching slower improves efficiency and helps on EMI issues but you only want to go so low. I liked 75kHz for a sub woofer application with a 100Hz type frequency repsonse. I set my LC filter in the low kHz range just because going lower meant a bigger physical size inductor and more cost but there is no reason you couldn't set the filter point at like 1kHz. You don't want to set the LC filter at 100Hz or too close as there will be phase shift as the filter begins to affect the signal. Use an input filter to nicely filter out the audio frequencies you want to filter. The LC filter is for filtering the PWM waveform. Or I don't know what I am talking about and ignore the above. -SL zenmasterbrian Class D design is complex. Slower sampling does give higer efficiency. I'd like to try and keep it bipolar. That was why I suggeste 25khz. Then I could put the filter in the khz range, as you said, because going lower means larger components. But still, some 25khz would beed through to the driver. BWRX Use a higher switching frequency to reduce the size and cost of the filter components. 100kHz would be a better starting point. Why stick with bipolars when you can use mosfets? zenmasterbrian I'd like to stay with bipolars, taking advantage of the fact that this amp doesn't have to cover the full audio range. I'd like to stay with bipolars because the same number of them handles more power than MOS. I know that such designs are complex and there are many trade off issues raised. Is it OK to let a significant amount of say 25khz, hit the driver? BWRX The next question would be what sort of power output are you looking for from the amp? You can always parallel a couple mosfets if you need more current capability. Higher frequencies going to the sub will just be wasted as heat in the voice coil. If the driver can handle it then it won't matter too much. zenmasterbrian Maybe 1000W. I'd like to try and stay bipolar. It seems impossible to totally filter out the switching frequency. So it has to be ultrasonic. Is there any agreed upon standard as to how high is high enough? BWRX 1kW is an awful lot of power to be aiming for if you don't really know what you're doing. (I'm only assuming this since you asked what a good switching frequency would be in the first place). At 1kW you'd need to supply a lot of base current to get bipolars conducting. You'd need a lot of current to charge up the gates of mosfets too, but it would only be for a short period of time. You'd have less losses too. It is impossible to totally filter out the switching frequency, but you can certainly reduce it to an acceptable level. There isn't any standard, but like I said before, most try to make the switching frequency reasonably high to reduce the size and cost of the filter components. kartino 1000W? At 4 ohm or 8 ohm? For new class D DIYer this is not an easy job. Further more you deal with high voltage switching, expensive parts and lot of of trial. On my brain the amps will need material such as mosfet like advances device ATP18XXXXXXX, with snubering, disable body diode (expensive diode too...) triangle generator... imho you could not successfull with self oscillate, gate driver ic. You can use half bridge topology for that power level. Don't use bipolar, you really need high grade mosfet. Hi zenmasterbrian, What bipolar for 1000W switching amps? Best regards, kartino zenmasterbrian Many switching amps go at high freq because they want the full audio range. I've decided that this is the lowest part of a multi amp system. So it doesn't have to be that high. But I do believe it needs to be ultrasonic, because it will not be possible to totally filter it out. Any good books that deal with class D. Usually it is considered beyond the scope of DIY, but still any thing good to read?:D fredos Sorry for you Kartino, APT have just been buy by MicroSemi and nothing is avaible for now, just wish that they will be back on the market as soon as possible! All my D-Amp work with APT product, and now stock is low! PS dont ever think to use bipolar for switching amplifier, you will lose more power to feed current to output device base than you will put out of amplifier! Fredos www.d-amp.com mzzj quote: Originally posted by fredos PS dont ever think to use bipolar for switching amplifier, you will lose more power to feed current to output device base than you will put out of amplifier! Fredos www.d-amp.com Just wish that Eva doesnt see your comment or else she will rip you apart :D zenmasterbrian I'm going to read the materials on your linked page carefully. It looks like those amps lack a power transformer. I'm not real comfortable with that. It also looks like they don't have much filter capacitance. So they are using the output transistors to in effect, also regulate the DC power. Maybe this is unavoidable, but I still think it should be with in bounds. As far as which works better, bipolar or MOS, I think it really depends on the switching frequency. That is why I started this thread, to look at the trade offs. kartino APT low stock? Not good for your amp indeed, Fredos?. Did you ever try IXIS? I dream to use one of APT or IXIS but believe or not both are not available in our area. Best we have are only IRFP. fredos I've lokking for IXYS...Unfortunately, my major issue was not the MOSFET, but the rectifier diodes! Only APT was doing what I need...Anyways, maybe I will use double MOSPEC rectifier... Fredos www.d-amp.com PS, When I talk about bipolar, I dont talk about IGBT, but standard bipolar transistor! Workhorse quote: Originally posted by zenmasterbrian I'd like to stay with bipolars because the same number of them handles more power than MOS. I know that such designs are complex and there are many trade off issues raised. Is it OK to let a significant amount of say 25khz, hit the driver? Compare APT30M85BVR with the Highest power Bipolar available and see which one has more power...... Yes Class-D designs were complex as long as you donot understand what you are designing..... Bipolars consume lot of Base current[Several amperes] and are Harder to Drive whereas Mosfets require only fraction of ampere to DO THE JOB! Hi Fredos, I emailed Them and they answered it : "Microsemi is due to release the power devices in the market as soon as they rectify some managerial issues within APT....." regards, K a n w a r classd4sure Gate driver current capability is perhaps relaxed in direct comparison to BJT's however it isn't at all to be neglected, in fact I'd consider it an area of concern. So Fred, what is it that APT does with the body diode that's so very unique.. and that can't be solved by other means? Arent' they extremely expensive as well? I know they make a great product, but very expensive as well. I'm sure you've explored the alternatives and weighed the options, why not share a little of that with us if you please. zenmasterbrian Whether MOS, IGBT, or BJT is best, really depends on the frequency. That is why I started this thread. Given that this is a subwoofer amp, the Nyquist frequency is going to be low. But I think it still needs to be ultrasonic, no? So, how low can it be? Then we can talk about how to realize it best. BWRX What minimum impedance will you want the amp to drive? zenmasterbrian Since it is for a subwoofer, and that could be an automotive driver with 2 4ohm coils, and since it will be at the speaker, and since it will be kelvin wired, Lets design for 2, and possible even 1 ohm! Genomerics quote: Originally posted by zenmasterbrian So, how low can it be? Then we can talk about how to realize it best. How do you manage personal relationships whilst being such a vacantly pretentious person? DNA mzzj quote: Originally posted by Workhorse Compare APT30M85BVR with the Highest power Bipolar available and see which one has more power...... Like 1-2kV and several kA capable bipolars? :D OK, fair enough, your mosfet has a little bit better switching charasteristics :D zenmasterbrian MOS still has higher on resistances than Bipolar, agreed? At a low switching freq, bipolar can still do more power with the same number of transistors and the same heat sink. The issue is, what frequency is the turning point. But before that, what frequency needs to be used here? That's why I started this thread this way. BWRX Paralleling mosfets is better than paralleling bipolar devices because mosfets have a positive temperature coeffiecient and bipolars have a negative temperature coefficient. Since mosfets perform better at higher switching frequencies than bipolars there's no reason they wouldn't be superior at lower switching frequencies as well. Bipolars are current gain devices so whatever advantage they may have in on state resistance will be lost in the required base drive circuitry. Just shoot for a switching frequency around 100kHz as a starting point. fredos Hi classd4sure APT for now is the only device that can handle the power output I need with only 4 output. I have made lot of test, even with IXYS and for now only APT is able to handle that power at that effiency...Something special here! Fredos www.d-amp.com zenmasterbrian Switching power supplies went to 40khz and 100khz, and beyond because they want the magnetics to be as small as possible. So their design requires MOSFETs. But for this, a Class D subwoofer amp, minimizing the size of the magnetics is not an issue. Rather, I'd like to get the highest efficiency possible. How high, depends on how low the switching frequency can be. Assume the amp never has to deal with anything greater than 200hz. So, for the Nyquist criterion, I just need 400hz, plus some filtering buffer. But in practice, it is had to filter at that low of frequencies. Also, there will still be some of the switching signal left. So I believe it has to be ultrasonic. Are there examples where people built class D and kept in in say the 22khz range to 30khz???? I don't want to irritate dogs. Is there some agreed upon minimum frequency for such things??? SpittinLLama Just use 100KHz like suggested and get on with it, 3 pages about this now and still no decision. Or design your circuit so you can adjust the frequency easily between say 50KHz and 100KHz and get going on some real design. -SL kartino OK Zen Master, we are waiting your 'big base driver' maybe we can use your design to drive mosfet too. :cheers: Anyway as I know so far, the winner for smallest 'on resistance' is still APT18_, do you have candidate from bipolar? zenmasterbrian I'd have to look in data books. But with a pair of TO-3 bipolars on large heat sinks, no fans, you can do 200 watts for recorded music. For PA or musical instrument that number will be lower. MOS will generate more heat for the same current flow. So you will need more of them. Wheter MOS works better or bipolar does, is largely a matter of frequency. MOS will be better above a certain frequency. So the place to start here is frequency. So, is there some standard for low high you have to be, so as not to bother dogs or cats, etc, with ultrasonic getting through a filter?:cool: Tim__x Frankly I don't think you have the knowledge to undertake such a project. quote: a pair of TO-3 bipolars ... 200 watts for recorded music. You seem to be confusing class B amps with class D. quote: MOS will generate more heat for the same current flow. The same current with the same voltage across the device will produce the same power dissapation no matter what type of transistor it is. P = V * I, device type doesn't enter into the equation. zenmasterbrian When I said 200 watts for a pair of T0-3 bipolars, I meant class B. I say this, because it is more than you could do with a pair of MOS transistors. MOS drops more voltage, so it generates more heat. At low enough frequencies, bipolar will get better efficiency in class D, than does MOS. Early switching power supplies were always bipolar. They went to higher frequencies to get the size of the magnetics down, ans so they have to use MOS. So the question is, how low can the switching frequency be for a Class D subwoofer amp? Is there a good book that addresses this?:cool: BWRX quote: Originally posted by zenmasterbrian MOS drops more voltage, so it generates more heat. Say you're using an IRFZ48 like RX5 did in his project. That will have an on state resistance of 18mohms with a Vgs of 10V and a drain current of 43A. That equates to a voltage drop across the fet of only 0.774V. And that's only a continuous rating. When switching it can obviously handle more peak current and the losses will be lower. Don't forget that you'll only need to charge the gate up once and hold it there to keep the fet conducting. Don't forget you'll need to supply continuous current to the base to keep the bjt conducting. With the lower gain of power bjts you may need an amp or more of base current to saturate a bjt with a 43A collector current. Now, what sort of power dissipation will a similar bipolar device have? Read through this if you haven't already: http://sound.westhost.com/articles/pwm.htm There IS a reason you don't see many class d amps using bipolar devices anymore. zenmasterbrian I'll read you materials, and any more you can supply. If you can supply a link for that transistor? Especially a book all about class D design. But I know you can do more power per transistor at low frequency with Bipolar. It is a vertical process, so the resistance is lower. A 40khz MOS switching power supply has better efficiency than a 100khz MOS switching power supply. A 25khz MOS switching power supply will have still better efficiency. And at that low a frequency, a Bipolar switching power supply will have even better efficiency. The move to higher frequency was to get smaller magnetics in switching power supplies, and to get good broad band audio. I don't care about smallest magnetics. And I just want a subwoofer amp. The question I am really posing is how low a frequency can I go? Some will get through the filtering. So it seems that it must be ultrasonic. But how high? It seems an environmental concern? Is 25khz high enough? Will that not be perceptible, even subliminally? And will it bother animals? ( if it bothers insects, I'll be pleased ) Lets address the frequency issue, then the implementation technology. zenmasterbrian BWRX, what power spec is on the data sheet for the transistor you mention? Understand that that power spec is just a measure of its thermal resistance. It is die size, the package, and the mounting substrate thickness, between the die and the package. It is also substrate material. For bipolars, in the Motorola catalog, the highest is 350W, NPN only. Then there is a pair of complementaries at 300W. But the breakdown voltage is really too low for audio. Then there are transistors at 250W, complementary. These are widely used for audio. Actually the TO-3 power rating is the same as for PowerTab. But! That will drop more for power tab if you have it insulated from the heatsink. These numbers are for the case being kept at 25 deg C, which is unrealistic. What is it for your best MOS? But also understand that with a switching, or Class D application, the power disipated in the output transistors is due to switching speed and on resistance. For on resistance, bipolar will be better. But lets first focus on frequency, how low can this be? Then look at technology. zenmasterbrian Actually, for Class D audio, that 350W Motorola transistor could be the way to go. Given that I'll probably be at 4 or 2 ohms, so I won't need a huge voltage, and given that Class D would not seem to benefit from any push pull or four transistor bridge, that 350w NPN could be the way. Some could be paralleled ( with small emitter resistors ) and run between a positive and negative supply, with the filter components. Maybe not. maybe still need two transistors. But anyway, I know class D will be much different from linear. First, is there any agreement on how low a frequency you can spew out in a driver, without being considered an environmental hazard???? zenmasterbrian I would love to see a book that analyzes all the aspects of class D audio. I suspect that two NPNs in the quaisi complementary config. is the best way. So, an emitter follower on top of a common emitter. But again, frequency? Say I have a 1000Watt amp. Say I've got a filter that takes the switching signal down 50dB!! Say this amp puts 0.1 watts on the driver at the switching freq any time it is turned on. Now, the driver is a stout subwoofer. It won't be bothered by this. Also, overall efficiency is not bothered by this. But is it a psycho acoustic problem?? Would it bother humans? How about birds and animals? If it was 0.1 watts at 25khz? Is that ok? 22khz ?? It makes a difference. From tables I've seen, there are efficiency drops when you start to go higher. Also, you need to go to MOS. This does not increase the efficiency, but it does stop it from being degraded as much. So the first question is how low can the switching frequency be?:cool: Workhorse Hi ZEN, What made you to think THAT MOSFETS PRODUCE MORE HEAT DISSIPATION THAN BIPOLARS.... regards, K a n w a r zenmasterbrian In a linear amp situation the disipation is not going to be higher. But still, the MOS devices are not usually able to disipate as much. Here, I am looking at Power Control with Solid-State Devices by Irving M. Gottlieb. He says that ON resistance of MOS is clearly higher than comparable size bipolars. So when used in a switching power supply or class D amp, the thermal disipation will be higher for MOS. Right now, I want to talk about frequency. According to Gottlieb, early switching power supplies ran at 20khz to 25khz, and were bipolar. They didn't want to run slower because of audible noise. The move to higher frequencyies was facilitated by MOS, but the purpose was smaller magnetics. MOS lets you get the higher frequency and smaller magnetics, but it still does have a higher on resistance. Clearly for a broadband, up to 20khz class D audio amp, you need to use more than 40Khz, and most probably will go MOS. But I am only interested in a subwoofer amp. So how low can I go, given that some of the switching frequency will get through the filter? 25khz? 24khz? 22khz? I don't want to create anykind of negative psychosubliminal effect. There must be some conventional wisdom on this. After this is resolved, we can look at semiconductor technology. BWRX quote: Originally posted by zenmasterbrian So how low can I go, given that some of the switching frequency will get through the filter? 25khz? 24khz? 22khz? There must be some conventional wisdom on this. It's called the nyquist criterion. The absolute minimum sampling frequency required is 2x the maximum frequency being sampled. quote: Originally posted by zenmasterbrian The move to higher frequencyies was facilitated by MOS, but the purpose was smaller magnetics. You make it sound like you want to use humongous inductors and capacitors??? Smaller magnetics are beneficial for a number of reasons. You can get higher quality components that are smaller and have less parasitics. They can obviously cost less too. Higher switching frequencies push the noise further away from the audio band and allow the use of lower value inductors and caps which again have reduced parasitics. If you're having trouble deciding on a switching frequency and think bipolars are better than mosfets how are you going to design a 1000W amp? Not bashing you or anything, more curious than anything. kartino In this age? BJT beat Mosfet? Thanks to Mr. Bardeen, Mr. Brattain and Mr. Shockley. Tim__x Because of base drive losses, BJTs can only compete in terms of efficiency in very high voltage low current applications where only a small current at ~0.7v is needed to switch a moderate power (high V, low I). In a class D amp driving 2 ohms a very large current will be required (again at ~0.7v) to switch a moderate power (high I, low V). Yes, BJT are used in switching power supplies, but you'll note 99% of the time they are used on the primary side. Have you ever seen a synchronous rectifier made with BJTs? zenmasterbrian These are many complex issues being raised. First, the Nyquist criterion is not an issue here. I only want a Subwoofer amp. So lets say it will never have to handle an audio signal of greater than 200hz. So, then 400hz would be the minimum? Problem is that some will get through the filters, so it will be audible. So the frequency question here is how low can I go and not have it be audible, subliminally noticable, or an environmental hazard for pets or wildlife? Second, how big a magnetics would I have to use? Say I had a fourth order filter at 1000hz? It would be comparable to a passive speaker crossover at 1000hz. There are people who make fancy foil wound inductors and all. Also, if this is well designed, the filtering can be inside the feedback loop. So the signal being delivered to the driver is the one that will be compared with the input. Early switching power supplies ran 20 to 25khz. This is still published as a frequency range for inverters where bipolars beat mos. No, I have never seen a synchronous rectifier using BJT. But part of the reason they us MOS is to take advantage of the parasitic diodes. These diodes help in most of the duty cycle. ( I have seen on paper synchronous rectifiers using BJT ) Base drive current for BJT used in switching applications may not be trivial, but it is not a key issue. The issue is thermal disipation in the output devices themselves. It it this thermal dissipation which makes such output devices heat up, and determines how much power you can get per device. This thermal disipation in the output devices is caused by two factors. First, the on resistance or internal voltage drop. Second, the switching times, for both parts of the cycle. So, at low freq, BJt is better. At high enough frequencies, MOS is better. Now days, some people are saying that for certain speeds, and certain voltages and currents, GTOs are better, as are IGBT. There are also a few other new devices too. But for this set up, its BJT or MOS. The transistion point is going to be somewhere around 25khz to 30khz. What ever technology is used, there will be less heat in the output transistors if the switching frequency is lowered. Ok. But I don't want to talk about output technology yet. I started this thread to talk about frequency. How low can I go, with out being a psychoacoustic problem?? Its likely to be around 27khz, 25khz, 22khz? I would like to go much lower, to get better thermal efficiency in the output devices. But that would be audible. There must be some conventional wisdom on this. BWRX quote: Originally posted by zenmasterbrian First, the Nyquist criterion is not an issue here. I only want a Subwoofer amp. So lets say it will never have to handle an audio signal of greater than 200hz. Why wouldn't it be an issue here? Comparing the audio signal to a periodic signal is sampling, and yes 400Hz would be the min. Some will get through the filters and be audible which is exactly why people use much higher switching frequencies!! quote: Originally posted by zenmasterbrian So the frequency question here is how low can I go and not have it be audible, subliminally noticable, or an environmental hazard for pets or wildlife? What's the highest frequnecy that can be detected by anything living? Take that and double it. djQUAN just to note, I'm already on the planning stage in building a class D sub amp doing 1kW into 1 ohm. I also want to lower the switching frequency but decided to use atleast 100kHz. more around 250kHz. :D I'm not trying to start anything here, but I'm curious......what's the point in asking in the first place if you don't want to accept the advice given to you? I have about 300pcs of switching transistors and still haven't gone to using one of them because of the rather high current needed to keep them conducting compared to mosfets. why did I get them in the first place? coz I got them real cheap. $6 for the whole bag. I can use them someday when I build a SMPS that runs on a low voltage suppy that doesn't have enough voltage to keep a mosfet conducting effectively (<5V supplies) mzzj quote: Originally posted by zenmasterbrian This thermal disipation in the output devices is caused by two factors. First, the on resistance or internal voltage drop. Second, the switching times, for both parts of the cycle. So, at low freq, BJt is better. At high enough frequencies, MOS is better. How does your jellybean pick bipolar compare to 5 or 10mOhm 100v mosfet in terms of saturation voltage? (100v rail is enough for 1kW to 4 ohms) Fast-switching bipolar tranny with less than 0.25 Vce sat at 50 amps and square switching SOA? SHOW THAT TO ME! Tim__x Conduction losses remain fairly constant with frequency, switching losses go up merely linearly with frequency. You say that the frequency where BJTs overtake MOSFETs is around 30K, it follows that at 30k losses are equal for the two device types, it also follows that, given the relationships mentioned earlier, at frequencies near 30k the losses are nearly the same. If you won't listen to us, pick whichever you prefer. As for switching frequency, as long as it's above 20k people won't hear it, dogs, cats, bats, mice and the sort might hear it up to 100k. Go for 50k or so, then put a 1k low-pass on it. With that, I leave this thread, having said all that is useful for me to have said. kartino The frequency is not the only issue. Even at slow frequency you still need fast ON and fast OFF to get real squarewave. Because here we talk about sound production that you don't want any distortion. While on switching power we can tolerate small distortion but the ear will hear even very small distortion. So if you are sure the BJT can do good squarewave then it is fine. For frequency normally for full range, say 500kHz signal carrier the audio sampled at least 25Hz per cycle so for 400Hz = 400 x 25 = 10.000Hz. Or if you still worry about audible you can use 25kHz to 30khz. And make suitable filtering. A big filter maybe. So please make a test for BJT performance and trial for frequency. zenmasterbrian That BJT out performs MOS for modest frequency switching is well documented. I got a book coming in which goes into this. If you have links for especially attractive MOS, lets see em. You might be talking about stuff much bigger than TO-3. Well, there are BJt bigger than TO-3 too. As far as distortion, remember that the filtering is inside the feedback loop. As far as sampling being adequate, 20khz is plenty for a 200hz audio signal. As far as filtering at 1000hz, fine. But some of the switching signal is going to get through. There must be some psychoacoustic data about what people and animals would percieve. I don't plan on doubling anything. Instead, I plan on using well what information is available. An amp for 200hz is not the same as a broad band audio amp. It is also not the same as a switching power supply. This minimum switching frequency requirement is external. It has to do with humans and animals. I'm willing to listen to anyone who actually knows anything about that. kartino Why you are worry about effect switching frequency for human and animal? There is no accoustic sound production since you use filter. Very easy to remove high frequency isn't? And remember that you use woofer instead of tweeter. Both you amp and speaker did not produce high audible ultrasonic. It is not make sense that so far you worry about creature injured by your amp for me. :dead: kartino And again feedback is not fix distortion if the feedback loop itself that produce distortion. kartino OK this is googling from me free of charge for you: http://www.lsu.edu/deafness/HearingRange.html Species Approximate Range (Hz): human 64-23,000 dog 67-45,000 cat 45-64,000 cow 23-35,000 horse 55-33,500 sheep 100-30,000 rabbit 360-42,000 rat 200-76,000 mouse 1,000-91,000 gerbil 100-60,000 guinea pig 54-50,000 hedgehog 250-45,000 raccoon 100-40,000 ferret 16-44,000 opossum 500-64,000 chinchilla 90-22,800 bat 2,000-110,000 beluga whale 1,000-123,000 elephant 16-12,000 porpoise 75-150,000 goldfish 20-3,000 catfish 50-4,000 tuna 50-1,100 bullfrog 100-3,000 tree frog 50-4,000 canary 250-8,000 parakeet 200-8,500 cockatiel 250-8,000 owl 200-12,000 chicken 125-2,000 BWRX quote: Originally posted by zenmasterbrian That BJT out performs MOS for modest frequency switching is well documented. This minimum switching frequency requirement is external. It has to do with humans and animals. I'm willing to listen to anyone who actually knows anything about that. Are you even reading anything being posted in this thread? I am utterly amazed how you continue to bypass all reaonable thought and end up going back to your original question which has been answered numerous times and is not dificult to answer in the first place. Why don't you build this bipolar based behemoth and tweak the switching frequency yourself to see what what works and what doesn't? zenmasterbrian kartino, than you for finding that reference. I'm still reviewing the article. It does seem that the hearing limit is fuzzy. I've decided on a tentative frequency specification. 23.5khz. Now, the amp only has to go to 300hz absolute maximum. Typically much lower. So there will be a filter. I've decided on 8th order. Remember, increasing the order of such filters does not increase the total reactance. Rather, it just distributes it. It will be comparable to a LP speaker crossover at 1000hz. The total amount of potential 23.5khz power will be about 2x the amps sine wave output power. But by 16khz it will be down 144dB. By 22.6khz it will be an extra 5th up, so it will be down 156dB. Now, I spent some time reading audio books last night, and reflecting on common experience. 1. 20khz audio propagates very poorly through air. It attenuates very very quickly. It is very directional. It won't do well at all with a subwoofer for a driver. It really can't even get through most grill cloths. Common cloth is very opaque to it. It is highly directional. So, I am not going to worry about outdoor animals and this system. 2. With all due respect to kartino's article, 20khz is a real stretch for human hearing. Even when it is there, the sensitivity is low. Before reading it I had picked 23.5khz, which I will explain more about. 3. I will plan on extra acoustic filtering in the form of a foam rubber grill cloth that is on a close fitting frame. 4. The filter, and much of the other circuitry will be well enclosed. Say wire fly screen, and then plexiglass on top of that. I still want to show this off. I will continue to look at any additional information about human and animal sensitivity to ultrasonics, as well as what precedents there are for appliances that might generate them. Again, back in the 70's, switching power supplies typically ran in the 20khz to 25khz range. Also, before I build this, I will experiment with a woofer and a signal generator, running the frequency way up. I will also experiment with the smallest of dome tweaters and piezo tweaters, going to 23.5. I will let other people and animals experience it. More about 23.5 khz in next post zenmasterbrian 23.5khz i say. but actually i want a range of 23.5 to 25.5. I want this to have an unusual and not strictly necessary feature of being able to synchronize, phase lock. Consider this an audiophiles purism. If there are any drawbacks to this time domain sampling, it can be minimized by such. So this amp could phase lock to another amp of the same design, or to a digital source. So, it is 23.5Khz absolute minimum, and when running free. If there are multiple amps in the system, you is switched into leader mode, at 24khz. If there is a DVD, it locks at the 24khz. If it is CD audio, then the 44.1 x 4/7 = 25.2khz. I know this is unnecessay, and not provided for in most consumer gear. I know it means extra wires and all. But I still want this capability. But this PLL is done so that never ever does it go below 23.5. Also there is a sentinel circuit that watches both the speaker terminals, and the input to the filter. If this thing slows for one cycle, the speaker relay opens, and the power mains locking relay is released. zenmasterbrian Audiophile stuff is extreme and personal. It is designed to be the best possible. It is designed for exactly what it does, not in imitation of other appiances. This is for a multi-amped system. This is not a broad band amp. This is not a compact, low cost, switching power supply for a computer, either. A pair of linear tweeter amps using mosfets, to go with this, is likely. zenmasterbrian Now, if anyone has any additional information about human or animal sensitivity to ultrasonics, I would be very interested in seeing it. Also, if anyone does want to post links to low resistance transitor data sheets, I would like to look at it. BJT, or MOS, or anything else. If anyone has info about such circuit designs and their issues, I would like to see it. If there are books especially good about class D, or even about switching power supplies, I would like to know. zenmasterbrian This amp is not going to be small. The subwoofer enclosure it is integrated into will not be small either. BWRX quote: Originally posted by zenmasterbrian So there will be a filter. I've decided on 8th order. Good luck with that! quote: Originally posted by zenmasterbrian 1. 20khz audio propagates very poorly through air. It attenuates very very quickly. It is very directional. It won't do well at all with a subwoofer for a driver. All very good points. quote: Originally posted by zenmasterbrian 2. With all due respect to kartino's article, 20khz is a real stretch for human hearing. Even when it is there, the sensitivity is low. My upper limit is around 17kHz. Amazingly enough my father's upper limit is 21kHz. He's 59. I know he's not lying either because he could tell me when I switched my tone generator on and off while not looking at me! Everybody's hearing is slightly different. zenmasterbrian I'm planning on two versions of this amp, small and big. The small one is for a movable subwoofer. So it plugs into 115VAC, and at power peaks will take about the most you can get out of a 15A outlet. This is about 180O Watts. So it will have an 1800 Watt toroidal power transformer. The steady state power is much much less. But the core has to be big enough for that 1800W. The large version is for a built-in subwoofer. This will have an even bigger transformer, which is hard wired to 230VAC. Its power will be far greater. As long as I leave the electric stove off, I've got 12,000 Watts of service to play with, although its for all the audio channels. ( At the San Jose Tech museum there is an earth quake simulator which has a 12KW or 20KW sound system! ) Part of the reason for wanting such extreme capabilites, is just purism. But I am also interested in the 64' pipe organ octave. I want to experiment with infrasonics. Specifically going down to 8hz, or maybe a bit lower. I'm not sure if this can really be used in music, but I want to experiment. There are three pipe organs in the world which will do the 8hz. I believe it could make for a very interesting kind of ambient music. This system will also make for superb bass for the more standard musical range. The 32' organ octave, going down to 16hz, sounds wonderful! A 32" kettle drum ( the most common large ones are 28", for 16' F ) will go down to 16' C, at 32hz. I once heard a cover of Niel Young's "Old Man Take a Look at Your Life" The song was in the key of D, and they used a pair of such kettle drums at C# and D. It sounded wonderful. I know that some stuff is on the market now that does not use power transformers. I have reservations about this. All of these matter I have broached in subwoofers->Giant Subwoofer. zenmasterbrian BWRX, how much signal strength do you need to have for your father to hear 21khz. And what are you using for a transducer? What do you believe is the lowest ultrasonic people can be exposed to on a continual basis, without being even subliminally effected? BWRX I was using a TA2021B based Tripath amp hooked up to my speakers which use a 1" titanium dome tweeter. They're 8 ohms and around 90dB @ 1W/1m. I had the amp turned up more than normal, but the Tripath amp is limited to around 8-9W. I don't know the answer to the ultrasonic frequency question. I do know a graduate student who was working with high power ultrasonic transducer arrays and he felt some pain in his ears while working around an array that was putting out about 140dB @ 1m. He didn't know it was on because it was set around 40kHz but he definitely felt its presence! He thinks he may have damaged his hearing... zenmasterbrian 140dB is quite a bit. Perhaps he was experienceing heating in his ears, or some other kind of tissue denigration? But I am concerned about the felt presence. I know that the very highest frequencies which can be heard, are experienced in this way. I know what i want to do. But I don't want it to be something which is a physiological hazard. The 8 or 9 watts on the tweater is quite a bit. A 1" dome tweeter will work at 21khz or more, but it will put out a beaming pattern. Straight on axis will be fine. That much power is a strong signal. I once made smoke come out of a pair of tweeters with sustained ultrasonic once. 8 or 9 watts continuously could over heat many tweeters. zenmasterbrian If you don't mind my asking, what was the 140dB 40khz for? How much power did it have, and what was being used as a transducer? BWRX There is a 12dB/octave crossover network and resistive divider between the amp and the tweeter. I'm sure I wasn't feeding it 9W but I'd bet it was putting out in the realm of 100dB @ 1m. The 140dB 40kHz was produced with a very efficient array of piezoelectric transducers - I think there were 16 of them on a clear plexiglass panel if I remember correctly. Not sure how much amplifier power was used and I don't know exactly what they were using the array for. At the time the grad student was an instructor for one of my acoustics classes and he was showing us some of what he was doing for his doctoral thesis. zenmasterbrian Thanks. 100dB at 21khz is still a very strong signal. For the more typical kind of class D amps, running at higher frequency, how many orders of filtering do they have, and starting at what frequency? As I see it, a very big subwoofer could approach 100 dB/ 1 meter. Say I had 2000 watts, that would make it 133 dB. But then I have 156dB of filtering. Ok, but that is a 1khz filter. Going out above 20khz, there are parasitic capacitances in the filter, so it might not continue to roll off that far out . So there could perhaps be some 30dB of signal remaining?? Maybe?? I still think that could not be heard, even at 20kHz, and especially with a subwoofer. But at 23.5 it really should be unhearable. But then I have my foam rubber grill cloth too. I believe there must be precedents of people making bipolar subwoofer amps. There must be precendents for the lower switching frequency. People must have done this. Even an MOS amp will be able to do more power per output device if the switching frequency can be lowered. BWRX quote: Originally posted by zenmasterbrian For the more typical kind of class D amps, running at higher frequency, how many orders of filtering do they have, and starting at what frequency? Going out above 20khz, there are parasitic capacitances in the filter, so it might not continue to roll off that far out. Most high power class d amps intended for audio use switch up in the 100-500kHz range and most all use a simple second order LC filter on the output. Parasitics don't have much effect up until the RF range for the smaller components. Larger componets will obviously have more parasitics which lower the frequency at which the component starts to stray from ideal operation. zenmasterbrian Using only a 2nd order filter, even with such high switching frequencies, will still result in significant switching signal going to the driver. Of course it is a very very high frequency. There must be precedent for subwoofer class D amps in the frequency range I have in mind. These issues must have been dealt with. BWRX quote: Originally posted by zenmasterbrian Using only a 2nd order filter, even with such high switching frequencies, will still result in significant switching signal going to the driver. Of course it is a very very high frequency. If you use a high enough switching frequency and design the LC filter to have a cutoff in the 30-50kHz region then there will be enough attenuation of the switching frequency to not worry about anything that does make it through. quote: Originally posted by zenmasterbrian There must be precedent for subwoofer class D amps in the frequency range I have in mind. These issues must have been dealt with. There is a precedent, but this precedent is what gave class d amps a bad rep in the first place! Many people still think class d amps are only suitable for subwoofer use because of effective marketing and a general misunderstanding of how class d works. zenmasterbrian If your filter is set to 30khz, so it is 3db down there, and it is 2nd order, Then at 120khz you will be 40dB down, and at 240khz, 60dB down. This amount of signal is not a serious hit on power efficiency, but it is significant. Rember it is a rail to rail square wave. So it is about double the amps rated sine wave power. Now, some of those upper harmonics will be more attenuated. What really drives this for me is not an interest in BJTs, in Class D per se, or a belief that Class D is only suitable for subwoofers. The first consideration is wanting to do the whole system multi-amped, with active crossovers. So, you end up having a subwoofer amp separate. For the tweeter amp, I don't see any reason to go class D, even though I'm quite convinced that it could work. So for the subwoofer I want class D, at a modest frequency. Have you ever seen any published info on such Class D subwoofer amps? I can manage the 8th order filter and closed loop issues. BWRX quote: Originally posted by zenmasterbrian If your filter is set to 30khz, so it is 3db down there, and it is 2nd order, Then at 120khz you will be 40dB down, and at 240khz, 60dB down. This amount of signal is not a serious hit on power efficiency, but it is significant. Rember it is a rail to rail square wave. So it is about double the amps rated sine wave power. Have you ever seen any published info on such Class D subwoofer amps? I can manage the 8th order filter and closed loop issues. An 8th order filter is pretty rediculous at the frequency and power you're looking at. Do you have any idea how large the coils will have to be to have low winding resistance? Then you want to put 4 of them between the output stage and the driver? I don't see any advantage in doing that. An LC filter has a 12dB/octave slope or 20dB/decade. For a filter with a cutoff frequency of 30kHz that would have attenuation figures of 12dB at 60kHz, 24dB at 120kHz, 36dB at 240kHz, 48dB at 480kHz, and so on. Sure some ultrasonic stuff gets through the filter, but it doesn't have anything to do with power efficiency. Just like 20kHz will not be reproduced by a subwoofer driver, the switching residual will not be reproduced by woofers, midranges, or tweeters (maybe if you have tweeters with real extended frequency response...). It will mainly be wasted as heat in the voice coil. What most people don't mention is that there may be some serious advantages to the high frequency residual. Have you ever heard of an interesting technique they used to get better performance from magnetic tape? They summed the audio signal with a high frequency signal before recording onto the magnetic tape. Thie high frequency bias signal reduced the magnetic memory effect of the tape. Something similar may be happening in the conductive paths subject to high frequency residual. Just a theory really, but I know other people who have pondered this as well. Have a look here: http://www.hypex.nl/ and check out the UCD700. I suggest you buy those modules instead of trying to build your own with BJTs. They can get 700W from only 3 pairs of mosfets and a simple LC filter with 92% efficiency. Not too shabby eh? zenmasterbrian The 8th order filter does not require bigger total reactances than the 2nd order filter. They are just broken up into smaller units. It is about the same as a 1khz speaker crossover. ( you mention speakers with 1" dome tweeters. That is about where their cross over probably is. ) As far as coil resistance being a problem, the loop gain works to divide that down. Again, because this only has to work to 300hz, I can compensate it so I can get incredible loop gain. I want to fully take advantage of what making it subwoofer only makes possible. It really is designed for that application. So it is all the advantages of the lower audio bandwidth on the closed loop dynamics, a lower switching frequency, and I hope, the advantages that bipolar offers at such frequency. I am aquainted with the high frequency bias used in magetic tape. That works because tape recording is nonlinear, and stores info by hysterysis. I am not aware of any benefits of high frequency in an audio system. That sounds like metaphysics to me. I will look at the MOSFET module link you show. I am always interested to see such links. As I see it, for 3 pairs of devices on good heat sinks, Class D, 700Watts is not impressive. I believe bipolar at modest frequency can do much better. But I will look. I know that 23.5khz will not be well reproduced as it goes into a subwoofer. But it is still a cause for concern if it is perceptable. I'm still hoping to find precedent for a subwoofer amp designed in the switching frequency range I speak of. All the better if it was designed with bipolar. BWRX quote: Originally posted by zenmasterbrian The 8th order filter does not require bigger total reactances than the 2nd order filter. They are just broken up into smaller units. It is about the same as a 1khz speaker crossover. ( you mention speakers with 1" dome tweeters. That is about where their cross over probably is.) It is not the same because you're looking to use the filter on the output of a 1kW amp. The crossover on my tweeter is around 2KHz. quote: Originally posted by zenmasterbrian I am aquainted with the high frequency bias used in magetic tape. That works because tape recording is nonlinear, and stores info by hysterysis. Precisely. You don't think there would be any benefit to having the magnetic fields of a conductor constantly changed by a high frequency bias signal? quote: Originally posted by zenmasterbrian As I see it, for 3 pairs of devices on good heat sinks, Class D, 700Watts is not impressive. I believe bipolar at modest frequency can do much better. 700W with 3 pairs of mosfets on a heatsink that small and you aren't impressed? You know that module is capable of 20Hz-20kHz right with low distortion? You clearly don't fully understand how difficult it is to accomplish such a task. zenmasterbrian The UDC web site is very interesting. I have started to read from it, and will continue. Clearly these are very well designed, very high quality, and very practicle devices. Help me know a little more. Am I seeing correctly that the big T shaped blue part is a heat sink, and then there are three power tab transistors on each side? If you had data sheets for these transistors, and anything else it uses, I would be interested. They also say you can use only partially filtered DC power. You almost have to allow that at these kinds of power levels. When you allow big power ripples, you are placing additional demands on the output devices, although this increase is less for class D than for linear. What seems to be the heat sink is not what I would consider to be a big heat sink. It is a practical heat sink. I'm glad to know about this UDC company. It looks like their things are well made and very interesting. Again, my starting point here is multiamping, and the opportuities afforded when each amp in the system can be designed differently.:) BWRX quote: Originally posted by zenmasterbrian If you had data sheets for these transistors, and anything else it uses, I would be interested. Buy one and you can look them up for yourself :) zenmasterbrian Mr. BWRX, we have never met. So we don't know each other. Online communication is slow and sometimes difficult. But I think it worth the effort. The 8th order filter that I would use does not have to be much bigger than what is in your speaker cross over, even though the power level is higher. The inductors would be air core. So there is not iron or ferrite core that could saturate if not big enough. So its just a matter of the wires being big enough to not over heat. That won't be a problem. Also my high power is a peak power. The physical size of the inductors will not need to be much bigger. Also, I could tollerate much more DC resistance than you passive speaker crossover could because I am inside a close loop, which will be well designed. That is something I have studied well and in excruciating detail. The capacitor merely needs to have a high enough voltage rating. Since it is not electrolytic, this will not be a problem. Remember, this amp will be part of a big subwoofer enclosure. The total capacitance and total inductance will be about the same as the speaker crossover parts. TO BE CONTINUED zenmasterbrian Clearly the UDC amp is well designed. I am impressed with the good performance all the way up to 20khz. For my application though, that frequency range is not imporant. I am impressed with how small it is, but that is also not an objective for my appilcation. So of this is the difference between hobby and DIY stuff, versus mass market stuff. Again, multi-amping is the premise I am starting with, not class D, or small and compact. BWRX You may call me Brian, Brian :) What value will the inductors in your output filter be? More importantly, what class d topology will your amplifier be? zenmasterbrian As far as the high frequency on the conductors? No, I don't go along with that. I will have some of that myself. If that was really desired, linear amps could be made to do that. The basic premise is that the wires are linear. If they are not, we are in trouble. Its not like magnetic recording tape. I beleive that some of this stuff is more metaphysics, and typical of what you see in "High End" Audio circles. Do you have one of these UDC amps? I'm sure it works very well. As far as heat sink size, I feel that what they have is practical. It is more of a thermal mass. I'm sure it is a practical way of meeting their objectives, small size being one. When I speak of a big heat sink, I mean one that lets you come closer to the data sheet disapation level. For home audio, I see this as being a large aluminum heat sink, positioned to promote convection. But also put the transistors on live copper carriers. Then the mica insulator surface can be about 3 square inches. Often these mica insulators are the biggest thermal resistance. I know that for may applications such a bulky approach would be undesired. Again, DIY audio is often done for different objectives. Big heat sinks are one of the things solid state can have to make it showy like tube amps. zenmasterbrian Brian, I haven't figured out the inductor values. I've designed filters before. I've made tables for filter design. I've written progams for filter design. This was all years ago. There would be four inductors and four capacitors. They will not be the same sized. But the total of all the four inductors and the total of all the capacitors would be about the same as a second order LP speaker cross over for the same impedance, at the same 1000hz. Some people sell foil wound inductors for speaker cross overs. But I can tollerate much more resistance because of the feedback loop, and because of its narrow bandwidth. Lets say for 4 ohms, the inductance would total 0.7 milli henrys. The capacitance would total 40uF. ( these values are sold in non-electrolytic from specialty audio sources for speaker cross overs. As far as class D topology, I assume you mean the output transistor configuration? I do not know . First I wanted to get the minimum frequency down, these physiological issues. I have only thought about this a bit. It will be influenced by switching power supplies. But this is different because I can't bias the speaker. Perhaps it would use two NPN transistors stacked, with the output to the filter between them. So the top one is an emitter follower, and the bottom one is common emitter. This is usually called quasi complementary. Maybe? The biggest bipolar I know of is in the Motorola Catalog. 350 Watt dissapation, in a TO-3 package. What this means is that if you could keep the case at 25 deg C, and were willing to tollerate 175 deg C junction temperatures, you could dissapate 350 Watts continuously. In practice, using the kind of big heat sinks I describe, you could dissapate about a 3rd of that continuously. Now, this is for music power, not steady power. So how much peak power it could do is very subjective. Best to have thermal shut offs. The biggest bipolars for linear audio are usually 250W each, rated in the same way. Two of them, on the kinds of heat sinks I describe, can do about 200W musical peaks. I would like to think that two of the 350 Watt bipolars I specify could do 1000W in switching, but only at modest frequency. Just for comparison, as I know, the high power MOS transistors used for linear audio are rated at lower dissapation. But the dissapation numbers are not the real issue when comparing MOS and Bipolar for class D. Then it is the on resistance, and the switching speed relative to the switching frequency being used. I know that many designs don't use the scale of heat sinks I am wanting here. That is their right, and often entirely appropriate to their overall design objectives. Transistors are often cheaper than heatsinks. Brian If you have links to big transistors, I'd like to see. zenmasterbrian At the end of the thread Chip Amps-> Chip amps with built in heat sinks, we are talking about such heat sinks. Tim__x I'm sorry, I said I wasn't going to post again, but this is too irrational for me to ignore. quote: As far as coil resistance being a problem, the loop gain works to divide that down. So you care about a few tens of milliohms of MOSFET Ron and it's effect on efficiency, but you don't care about potentially hundreds of milliohms of inductor resistance in series with your load? :cannotbe: :headbash: Please, don't just assume I'm wrong or out to get you. Think it through carefully and you will see the inconsistency in your argument. Workhorse :devilr: zenmasterbrian I'm sorry, but I think some of you don't have detailed electronic design experience. Thats OK. But look, lighten up. MOSFETs definitely have higher resistance. The resistance in these devices is what causes the heat in them, and that is what limits how much power an amp can be designed for. I'm not concerned about this resistance because of over all power efficiency. I'm concerned about it because I want to design an amp with lots of power, and this resistance is what gives the output transistors a thermally based power limit. classd4sure Hmmm, I grow confused, ZenMasterBrian. The world is telling you you're in error. You're not listening. The world can't be right, so, you must know something the world doesn't? How about a comprehensive pro /con list to further your points on the matter? That way we can pick it apart one by one for you on a point form basis all neat and tidy. Then again who's to say if you did make the list you'd still post it? BWRX quote: Originally posted by zenmasterbrian MOSFETs definitely have higher resistance. The resistance in these devices is what causes the heat in them, and that is what limits how much power an amp can be designed for. In my textbook entitled "Power Electronics" authored by Mohan, Undeland, and Robbins is a lot of good information about power electronic devices as well as converters, applications, and design information. Some good bits of info about power BJTs intended for power swith-mode applications: "Nearly all the power dissipated in the switch-mode operation of a BJT occurs when the transistor is in the on state. In this circumstance the power dissipation is given by (ignoring base current losses) Pon=Ic*Vcesat. The collector-emitter saturation voltage increases with increasing collector current." "Power BJTs have low current gain, especially at larger breakdown voltage ratings." "The SOAs of the BJT are limited by second breakdown." Thus, for BJTs with large voltage breakdown ratings (like you would use in your amp, Brian) beta is low. This means for large collector currents you will also need large base currents. This means you have lot more power dissipation in the base drive circuitry. Further, the negative temperature coefficient of BJTs means they are not as easily paralleled. Some good bits of info about power MOSFETs intended for power switch-mode applications: "Except at higher switching frequencies, nearly all of the power dissipated in a MOSFET in a switch-mode power application occurs when the device is in the on state. The instantaneous power dissipation in the on state of the MOSFET is given by Pon=Id*Id*Rdson." "The on state resistance increases significantly with increasing junction temperature." "On state losses in a MOSFET rise much more rapidly with blocking voltage rating than do those in a BJT." "The SOA of a MOSFET for switch-mode application is large (rectangular) because it is not subject to second breakdown." Power MOSFETs usually have higher Vgs ratings and require a larger gate-source voltage differential to fully turn them on and reduce Rdson. However, a large sustained current is not required to keep the MOSFET turned on. The gate capacitances need to be charged once and held at that voltage to keep the MOSFET on. This means reduced power dissipation in the gate drive circuitry. Further, the positive temperature coefficient of MOSFETs means they can be easily paralleled. Both devices have their appeals and turn-offs, but I don't see how you can overlook the use of multiple MOSFETs instead of a single or even multiple BJTs even for low switching frequencies. djQUAN I can see where this is going................ you explain, he ignores it and tells what he's been telling since the start of this thread. p.s: I'm not trying to start anything here. it's just what I've seen since I've been following this thread...... classd4sure Ditto. Also if we fall pray to arguing everything arguable we just end up doing all his research for him... just looks like a fisherman to me. BWRX I don't mind doing some research cause it helps refresh my memory as well :) I never said I was gonna help design the amp! On the contrary, I have absolutely zero interest in a 1kW class d amp using BJTs. I just figured it may sound more credible if I got the information from a textbook. classd4sure Hey I'm with you. Moments ago I read 98% of that plus another 10% other (comments on bandwidth/gm/noise too), just on some website that google found virtually on its own, as one of the first three links. Anyway he's been told all anyone can be told, bring on those designs I say :) mzzj quote: Originally posted by zenmasterbrian I'm sorry, but I think some of you don't have detailed electronic design experience. Thats OK. I think that only one in this thread without "detailed electronic design experience" is YOU. Have you ever designed and implemented SMPS or class-d amp? And its NOT OK because you are damn stubborn and dont listen anyone else than your own ramblings :devilr: :whazzat: zenmasterbrian classd4sure, Brian ( BRWX ), djQUAN, mzzj, I appreciate the fact that you all have different POV's. diy projects are going to be different then mass marketed products. But there is something you need to understand. NO ONE IS REQUIRED TO POST! Meta is contagious, and it drives legitimate posters away. I'm not going to read or respond to posts that amount to negative meta-conversation. If someone posts topical, then I will find it and respond to it. The rest of it I am going to ignore completely. I'm convinced that this board is full of good people, and so I encourage everyone to ignore all meta. Give me a little while now, and I'll try to respond to some of the topical portions. Also, remember that this thread is titled Class D frequency. It is a question about how low the frequency can be. I'm going to try and steer it back to that. I know that most class D designs are done for a different set of objectives. classd4sure Actually, it's how low it can be before you _need_ mosfets, seems to me the easy answer is build one and tell us. zenmasterbrian Regarding Mohan, Undeland, and Robbins: First, this sounds like a power supply book. So it is usefull, but not a prescription. Power supplys are designed for a different set of objectives than many diyaudio projects. Power supplies are usually designed for size and cost. Diyaudio can be different. All the admonishions about how hard it is to design with BJT are true. They go back decades and decades. Thermal runaway, current hogging, and high drive currents are all difficult to contend with. But also understand, that power and heat generated in drive circuitry does not contribute to the thermal power limit of the output devices. << Both devices have their appeals and turn-offs, but I don't see how you can overlook the use of multiple MOSFETs instead of a single or even multiple BJTs even for low switching frequencies. >> I think this is the most important statement from this text book. Both devices have their appeals and turn-offs. These authors do not see how you can overlook the use of multiple MOSFETs insead of a single, or even multiple BJTs, even for low switching frequencies. This is their opinion, and it is given regarding the design of power supplies for commercial markets. DIYaudio does not have to compete in that kind of a market. Such commecial market power supplies do not tend towards low frequencies, because they want the size and cost of the magnetics to be low. So the point of these authors is more a hypothetical than something that actually occurs. Further, I am not overlooking anything. Remember the title of this thread is Class D Frequency, not some BJT vs MOS debate. How low the frequency can be, will decide whether or not BJT can even be considered. In the statement made by these authors I do find an implicit admission that BJT has less voltage drop and therefore lower heat generation, and therefore it can put out more power per device. The key point for these authors would seem to be that you can parallel MOS devices easily, and get what you would have gotten from a smaller number of BJTs. That is true. I do not contest that. But again, they are still speaking from the realm of power supplies, where the lower switching frequencies are only a hypothetical, because of concern about cost and size of magnetics. What I am interested in is not a power supply, it is not a broad band audio amp, and it is not intended to compete on any kind of mass market. So the first thing to determine to see if what I want is viable, is how low can the switching frequency be? Seems like it has to be ultrasonic. But still what is the number? I have tentatively decided on 23.5khz, but I am still not totally sure that that is acceptable. I believe there are precedents of people designing class D amps that ran at that rate, in decades past. I belive there is some body of knowledge about what is perceptible. I have tried looking in different places. If 23.5khz looks like a go, I am likely to start a new thread to explore circuit toplogy and device technology issues. If anyone has specific links to transistor data sheets, for any technology, I would be interested to see them. All the links provided here have already been highly instructive. BWRX quote: Originally posted by zenmasterbrian First, this sounds like a power supply book. So it is usefull, but not a prescription. Power supplys are designed for a different set of objectives than many diyaudio projects. Power supplies are usually designed for size and cost. Diyaudio can be different. But also understand, that power and heat generated in drive circuitry does not contribute to the thermal power limit of the output devices. << Both devices have their appeals and turn-offs, but I don't see how you can overlook the use of multiple MOSFETs instead of a single or even multiple BJTs even for low switching frequencies. >> I think this is the most important statement from this text book. Both devices have their appeals and turn-offs. It's a power electronics book. The subtitle is "Converters, Applications, and Design", all priniciples which can just as easily be applied to class d audio amplifier design. After all that's what an audio amplifier is - a power converter. The heat dissipated in the drive circuitry doesn't contribute to the thermal power limit of the devices, but the base drive current of a BJT does. Also, overall efficiency is effected in a negative way with large base drive requirements which is an argument you used in favor of BJTs. That last <<snip>> is not a quote from the book. I did not put it in quotes because it is my writing. I know this isn't a debate over BJTs versus MOSFETs but output device selection is so important because you have to design the entire output stage around them. You can use MOSFETs at any frequency and expand the limits fairly easily by just adding more in parallel as long as the gate drive circuitry can handle it. You can't just do that with BJTs, but for whatever switching frequency you choose you should most certainly be able to use BJTs. Go to the manufacturers web sites and search for BJT datasheets. You don't need us to tell you which switching frequeny to use because 1) you won't listen and 2) it seems you've already decided on one. zenmasterbrian quote: Originally posted by BWRX You don't need us to tell you which switching frequeny to use because 1) you won't listen and 2) it seems you've already decided on one. Let me just clarify a couple of points. A class D amp is similar to a switching power supply. But it is not necessarily designed to meet the same contraints. Small and cheap is not necessarily an objective. Second, I am only secondarily concerned with power efficiency from the power mains to the speaker terminals. The efficiency I am really concerned with is in the output devices themselves, and that is only because that is what limits how much power I can design for. I've got an interlibrary loan request to get something that I remember has a table, showing that BJT is still used in some industrial power switching applications. I listen to everything. It is true that I have already decided what approach I want to explore the feasibility of. If it is not feasible, I will abandon it. But nothing I have heard so far has convinced me of that. I don't know what the environmental issues of low ultrasonics are. If there really are problems, that would sink my initial approach to this. Has anyone seen amps that run in that frequency range? I believe they have existed. Any links? Any links to material written about low ultrasonics pertaining to switching. I know that in the 1970's, switching power supplies did run in the 20khz to 25khz range. They did not go lower because of audible sound. mzzj quote: Originally posted by zenmasterbrian In the statement made by these authors I do find an implicit admission that BJT has less voltage drop and therefore lower heat generation, and therefore it can put out more power per device. IRFB4410 is a 100v mosfet in to-220 case. 8mOhms rdson typically, 10mohms max. BUV20 is one of the best bipolars that I can find. Its a high-current NPN bipolar in bigger and more expensive TO-3 case. VCE sat 0.3v typical and 0.6v max at 25amps. 4 times cheaper mosfet in smaller package have lower voltage drop.... (to220 vs. TO-3) TO220 bipolars dont have any change against TO220 mosfets so I didnt even take them to comparision. classd4sure I'd only make the single request that you don't go starting new thread after new thread and keep it all localized here..... please. Since you already answered your own questions it really is time to start seeing some circuits don't you think? At least lay out your design criteria. What have you set forth so far, slowest, most inefficient class d amp ever, for bass only. The thought occurred to me, perhaps you'd consider limiting your search timeline to everything from at least over a decade ago. That'll k