feel like i'm discussing amplfiers with Andy Capp 
If you're not after high output voltage (42V), but high output current (1R load), there's no need to use high power devices.
Especially not high Vce + high power devices (the MJL4281 is a Plastic critter btw, TO-3P)
Instead, you could go smaller, but many.
Take a look at the Thagard A75, Mr Thagard clocked his version at 70 amps out in 0R1.
An amp I built in the early '90s with Vgs-matched IRF630/9630 and lower value source resistors (quads of 1/2W metalfilms, likely stole that from Mr Pavel Dudek, some of his designs were published in magazines here a long time ago)
That is affordable (afair, I paid ~$150 for 200 MOSFET pairs), and spreading output current between more devices leads to less voltage error.
But as Mr Abraxalito said, even more devices in parallel is more Ccb.
For this you can take a look at the Thagard A100, has a cascoded output stage (TO-3 Motorola + TO-220 vertical MOSFET)
3-part article series in Audio magazine, thing I built in '95/'96 with regulated rails (both front end and current stage)
Copies of the article have been posted at diyA, btw.
A serial manufacture example of power amplifiers with cascoded output stages is Array (M-10, S-10), stacked Zetex SOT-223 devices (their reason for cascoding tiny devices is linearity/bandwidth).
Part of your problem are the rails, voltage sag (plus ripple for tens of amps out in 1R)
You'd have to regulate the rails, or do as Mr Wurcer suggested, use batteries.
But if you like class A so much, and have no issue with car batteries, I suggest Glen Kleinschmidt's idea (1024W RMS Class A+ thread)
With Glen's contraption, you could have 100W continuous class A power from 1-8 ohm. (without the silly amount of dissipation)
Wayne, here's a workable model.
I won't go into much lenghty explaining, just a basic note. The topology is what I call a Hitachi bastard topology because it looks a lot like what Hitachi serves us in their HMA-7500 MkII MOSFET amp. FYI, that topology has since beome one of the standard topologies anyone talking about them will evetually mentio, and indeed, Bob Cordell makes a note of it. I say "bastard" because it generally looks like that, but there are several changes which do change the original into something different even though similar.
The one thing I felt obliged to do is to use an unusually elaborate front end CCS. There is no current mirror, which makes the front end have a lower Common Mode Rejection Ratio (CMRR) than I'm comfortable with. Some months ago, somebody posted here 10 or 11 types of CCS circuits with measurements of their CMRR factor, and showed that this can vary from a humble number of 42 dB to a much better number of 82 dB, no less than 100 times better. The version I used was not on that list, but it was clear enough that if a cascode circuit was used, the rejection would be much improved over the more common types.
As you can see from the attached schematic, as per your wishes (which I still feel are risky) all overvoltage and overcurrent protection has been removed, leaving only fuses for protection. I'd never do that for myself, but you wanted it just so.
As noted in a previous message, a rough calculation showed you culd get away with 7 pairs, so that's what i used, 7 pairs of 200W plastic pack devices. I have to say that as is, it is running at full power at the very limit of what the transistors can take for any reasonable period of time, as they will likely overheat and trip the overheat protection, which I did leave in because I don't want the transistors melting on you, nor will I ever deliver anything that has no DC protection. I cannot, I have seen in my own room what happens when something goes wrong and a burst of serious DC passes through, turning the bass driver into Kentucky Fried.
Note that amp is AC. It can be developed into a DC amp, all the more so because it has a FET input. Speaking of which, you are probably not going to need C20 if you are sure your preamp blocks DC, simply leave it out and connect the two holes with a bit of wire.
Also, note that in real life, you will need TWO output relays working in parallel, as one will limit you to 16A steady state, although it should pass aroud 24A in short term peaks.
As it turned out, the number of compromises I had to make was way less than I expected. Yes, I had to use more global NFB, but it's still around 31 dB only, still WAY below that "you can't have too much NFB" school limit.
But I did get what you wanted, reasonably low THD and IM specs even into 1 Ohm load. Here's a list of preliminary measurements, with input filter limiting at 200 kHz:
Open loop full power bandwidth - 42 kHz, THD 1.3% at 42 kHz
Open loop full power bandwidth at 20 kHz - output voltage 18Vrms, THD 0,8%
31 mV produces 20Vrms output open loop, OL gain 645:1 (56.2 dB)
1.1V produces 20Vrms closed loop, closed loop gain 25.2 dB
Output 1W/1 Ohms, 1 kHz - THD 0.003%
Output 20Vrms, 20-20,000 Hz, 1 Ohm - < 0.06%
Output 20Vrms, 20-20.000 Hz, 2 Ohms - < 0.05%
Output 20Vrms, 20-20.000 Hz, 3 Ohms - < 0.03%
Output 20Vrms, 20-20.000 Hz, 4 Ohms - < 0.02%
Output 20Vrms, 20-20.000 Hz, 6 Ohms - < 0.015%
Output 20Vrms, 20-20.000 Hz, 8 Ohms - < 0.012%
Obviously, by 4 Ohms, we are in fact reaching the limits of the topology.
Mote than this, the square wave perofrmance looks really good, controlled overshoots, no ringing to speak of.
Even more important, harmonic decay looks very promising as well, but still needs some work to meake it really good.
If you want to discuss it further, I suggest we switch to private post, so as not to throttle the whole forum with this one-off product.
FYI, I am chalking you up for an extra small bottle of an alternative chilli sauce, but one which must make me run on afterburners.
PS. I forgot - the bias current is 115 mA per transistor, currently a total of (7 x 115) 805 mA.
I won't go into much lenghty explaining, just a basic note. The topology is what I call a Hitachi bastard topology because it looks a lot like what Hitachi serves us in their HMA-7500 MkII MOSFET amp. FYI, that topology has since beome one of the standard topologies anyone talking about them will evetually mentio, and indeed, Bob Cordell makes a note of it. I say "bastard" because it generally looks like that, but there are several changes which do change the original into something different even though similar.
The one thing I felt obliged to do is to use an unusually elaborate front end CCS. There is no current mirror, which makes the front end have a lower Common Mode Rejection Ratio (CMRR) than I'm comfortable with. Some months ago, somebody posted here 10 or 11 types of CCS circuits with measurements of their CMRR factor, and showed that this can vary from a humble number of 42 dB to a much better number of 82 dB, no less than 100 times better. The version I used was not on that list, but it was clear enough that if a cascode circuit was used, the rejection would be much improved over the more common types.
As you can see from the attached schematic, as per your wishes (which I still feel are risky) all overvoltage and overcurrent protection has been removed, leaving only fuses for protection. I'd never do that for myself, but you wanted it just so.
As noted in a previous message, a rough calculation showed you culd get away with 7 pairs, so that's what i used, 7 pairs of 200W plastic pack devices. I have to say that as is, it is running at full power at the very limit of what the transistors can take for any reasonable period of time, as they will likely overheat and trip the overheat protection, which I did leave in because I don't want the transistors melting on you, nor will I ever deliver anything that has no DC protection. I cannot, I have seen in my own room what happens when something goes wrong and a burst of serious DC passes through, turning the bass driver into Kentucky Fried.
Note that amp is AC. It can be developed into a DC amp, all the more so because it has a FET input. Speaking of which, you are probably not going to need C20 if you are sure your preamp blocks DC, simply leave it out and connect the two holes with a bit of wire.
Also, note that in real life, you will need TWO output relays working in parallel, as one will limit you to 16A steady state, although it should pass aroud 24A in short term peaks.
As it turned out, the number of compromises I had to make was way less than I expected. Yes, I had to use more global NFB, but it's still around 31 dB only, still WAY below that "you can't have too much NFB" school limit.
But I did get what you wanted, reasonably low THD and IM specs even into 1 Ohm load. Here's a list of preliminary measurements, with input filter limiting at 200 kHz:
Open loop full power bandwidth - 42 kHz, THD 1.3% at 42 kHz
Open loop full power bandwidth at 20 kHz - output voltage 18Vrms, THD 0,8%
31 mV produces 20Vrms output open loop, OL gain 645:1 (56.2 dB)
1.1V produces 20Vrms closed loop, closed loop gain 25.2 dB
Output 1W/1 Ohms, 1 kHz - THD 0.003%
Output 20Vrms, 20-20,000 Hz, 1 Ohm - < 0.06%
Output 20Vrms, 20-20.000 Hz, 2 Ohms - < 0.05%
Output 20Vrms, 20-20.000 Hz, 3 Ohms - < 0.03%
Output 20Vrms, 20-20.000 Hz, 4 Ohms - < 0.02%
Output 20Vrms, 20-20.000 Hz, 6 Ohms - < 0.015%
Output 20Vrms, 20-20.000 Hz, 8 Ohms - < 0.012%
Obviously, by 4 Ohms, we are in fact reaching the limits of the topology.
Mote than this, the square wave perofrmance looks really good, controlled overshoots, no ringing to speak of.
Even more important, harmonic decay looks very promising as well, but still needs some work to meake it really good.
If you want to discuss it further, I suggest we switch to private post, so as not to throttle the whole forum with this one-off product.
FYI, I am chalking you up for an extra small bottle of an alternative chilli sauce, but one which must make me run on afterburners.
PS. I forgot - the bias current is 115 mA per transistor, currently a total of (7 x 115) 805 mA.
Last edited:
Jacco, just two things.
Regarding the Motorola power devices, you just made the same mistake I did, you inserted a "J" which wasn't there. It's not MJL, it's just MJ. These devices were originally made as custom devices for Krell, but have since appeared as regular Motorola parts.
Just like you have say MJL 21193 in TOP-204 package and MJ 21193 in metal TO-3 package.
And secondly, since when is +/-42 V of PSU lines high voltage? True, Wayne did leave some doubt with his last report on required output voltage, he just said "17V". Is that 17V peak-to-peak, or is that +/-17V peak? Not exactly the same thing, is it?
Either way, I figured +/-42V full load on voltage will require 1,000 kVA toroids, one per channel. That's the price of having 1 Ohm speakers.
Also, I don't dare even starting to work out the heat sink requirement. I wouldn't be too surprised if it turned out that wayne needs to have two power amp towers simply because the heat sinks wll need to be like 1 m high each. They may well need wheels to be moved about due to sheer weight, because even using quick'n'dirty mental maths, I figure the heat siks will weigh in at something like 30 kilos, and each toroid, as per local manufacturer data, will also weigh in at around 8-10 kilos. So, he hasn't even started yet and already he's looking at over 40 kilos per side, likey to and up as 60 kg or more when fully asembled.
And I would imagine that if his good wife discovered that each one of those boxes was so heavy that it can't be moved if it has no wheels, hurricane Catarina will look like a picnic in the park for ol' Wayne.
Regarding your other suggestions, I would also strongly advise Wayne to consider them, although I think you'll find that in the end they would require so many devices that they would in fact be impractical to build. Imagine the number of output transistor C you'd have to have in parallel.
Last edited:
I will email you .....
Wayne, here's a workable model.
I won't go into much lenghty explaining, just a basic note. The topology is what I call a Hitachi bastard topology because it looks a lot like what Hitachi serves us in their HMA-7500 MkII MOSFET amp. FYI, that topology has since beome one of the standard topologies anyone talking about them will evetually mentio, and indeed, Bob Cordell makes a note of it. I say "bastard" because it generally looks like that, but there are several changes which do change the original into something different even though similar.
The one thing I felt obliged to do is to use an unusually elaborate front end CCS. There is no current mirror, which makes the front end have a lower Common Mode Rejection Ratio (CMRR) than I'm comfortable with. Some months ago, somebody posted here 10 or 11 types of CCS circuits with measurements of their CMRR factor, and showed that this can vary from a humble number of 42 dB to a much better number of 82 dB, no less than 100 times better. The version I used was not on that list, but it was clear enough that if a cascode circuit was used, the rejection would be much improved over the more common types.
As you can see from the attached schematic, as per your wishes (which I still feel are risky) all overvoltage and overcurrent protection has been removed, leaving only fuses for protection. I'd never do that for myself, but you wanted it just so.
As noted in a previous message, a rough calculation showed you culd get away with 7 pairs, so that's what i used, 7 pairs of 200W plastic pack devices. I have to say that as is, it is running at full power at the very limit of what the transistors can take for any reasonable period of time, as they will likely overheat and trip the overheat protection, which I did leave in because I don't want the transistors melting on you, nor will I ever deliver anything that has no DC protection. I cannot, I have seen in my own room what happens when something goes wrong and a burst of serious DC passes through, turning the bass driver into Kentucky Fried.
Note that amp is AC. It can be developed into a DC amp, all the more so because it has a FET input. Speaking of which, you are probably not going to need C20 if you are sure your preamp blocks DC, simply leave it out and connect the two holes with a bit of wire.
Also, note that in real life, you will need TWO output relays working in parallel, as one will limit you to 16A steady state, although it should pass aroud 24A in short term peaks.
As it turned out, the number of compromises I had to make was way less than I expected. Yes, I had to use more global NFB, but it's still around 31 dB only, still WAY below that "you can't have too much NFB" school limit.
But I did get what you wanted, reasonably low THD and IM specs even into 1 Ohm load. Here's a list of preliminary measurements, with input filter limiting at 200 kHz:
Open loop full power bandwidth - 42 kHz, THD 1.3% at 42 kHz
Open loop full power bandwidth at 20 kHz - output voltage 18Vrms, THD 0,8%
31 mV produces 20Vrms output open loop, OL gain 645:1 (56.2 dB)
1.1V produces 20Vrms closed loop, closed loop gain 25.2 dB
Output 1W/1 Ohms, 1 kHz - THD 0.003%
Output 20Vrms, 20-20,000 Hz, 1 Ohm - < 0.06%
Output 20Vrms, 20-20.000 Hz, 2 Ohms - < 0.05%
Output 20Vrms, 20-20.000 Hz, 3 Ohms - < 0.03%
Output 20Vrms, 20-20.000 Hz, 4 Ohms - < 0.02%
Output 20Vrms, 20-20.000 Hz, 6 Ohms - < 0.015%
Output 20Vrms, 20-20.000 Hz, 8 Ohms - < 0.012%
Obviously, by 4 Ohms, we are in fact reaching the limits of the topology.
Mote than this, the square wave perofrmance looks really good, controlled overshoots, no ringing to speak of.
Even more important, harmonic decay looks very promising as well, but still needs some work to meake it really good.
If you want to discuss it further, I suggest we switch to private post, so as not to throttle the whole forum with this one-off product.
FYI, I am chalking you up for an extra small bottle of an alternative chilli sauce, but one which must make me run on afterburners.
PS. I forgot - the bias current is 115 mA per transistor, currently a total of (7 x 115) 805 mA.
I had a Krell FPB300CX , it no likey , shuts off when pushed, Krell really cut back on the substance and increased the profit margins when they went to this model . Earlier 300 watt/ch krells had 5k transformers , this one had a 1.7k ....
Wayne, please remember that a nominally 1 kVA transformer delivering enough juice for 300W/8 Ohms will need to deliver at least 55-0-55 V at the secondary. That's 1,000/(55+55) approximately 9.1 Amps. Your 1,700VA trafos were good for 1,700/(55+55) 15.45 Amps.
In your case, that same trafo has to deliver just 30-0-30 V (full load on), so it should be capable of 1,000/(30+30) 16.7 Amps.
Locally made transformers go up in increments of 1,000 VA, so your next step could be a 2,000 VA trafo, which would double the current to 33.5 Amps. That is more than you want at the output worst case, and that's totally ignoring the rectification and filter cap effects.
As for the rectification, I'd use a full wave 250V/25A bridge rectifier for each supply line, 2 per channel, for the current gain stages alone. One of the reasons is purely practical - they are metal cased and thus very convenient for cooling, and make no mistake, they will need serious cooling.
20A into 1 Ohm is 400 Watts into 1 Ohm. Let's be extremnely pessimistic here and say you will have to drive a wild 1 Ohm load, and will need all of 2 Joules per every 10W dissipated (this varies from 1 to 2 Joules, depending on load complexity). So you will need 40 Joules. To get it, you will need 2x22,000uF per every supply line, or 44,000 uF per side. That's the minimum. Personally, I'd put two 22,000 uF caps in parallel, then add another 10,000 uF cap, and yet another 4,700 uF cap, to improve the speed of discharge. Plus all the stuff I drew in. That would give you 58,700 uF per line per channel (or 117,400 uF per channel), so assuming your full power on supply voltage does not go below 42V, you're good for about 52 Joules, a solid 30% above what you named as your maximum requirement.
We can play with this many ways, for example, Cornell-Dubilier make some very good quality 47,000 uF/50V milk can capacitors, which could stand in for the two 22,000 uF caps. But I would urge you follow the large milk bottle caps with smaller ones, as this improves speed of operation. Not at all all the same when you are dealing with 20 Amps of current.
I stress this because I noted from your messages that you tend to overdo the capacitor gig; while no saving can be tolerated in that department, overdoing will eventually start to counter productive, because the larger they are, the longer it takes to charge and discharge them.
You could, of course, use fully regulated power supplies. While possible in absolute terms, and while no doubt the best overall solution, you must remember two things:
1. Using electronic regulation in no way removes the need for large caps as it it wasn't there, thus skyrocketing the price and complexity, albeit with a good payoff if you get it right, and
2. Electronically regulated power supplies are a bitch to make because they must be unconditionally stable and literally lightning fast to be of any true use. This is not a lab stabilizied power supply, which is expected to deal with mA changes of current requirements, we are talking about a whopper here. In your case, electronic regulation would effectively be another power amp, one dedicated to DC opeartion, but still a power amp. Off hand, I'd predict you's need at the very least 5 pairs of the same output trannies, and probably more. Point is, it quickly gets out of hand.
- Status
- Not open for further replies.