OK
Now for a few latest waveforms...
The top left is a 300KHz
square wave at 10Vp per phase. bottom left is a 400KHz sine wave, Ch 2 is inverted and added to Ch 1 so this is what you see across the speaker. Top right is 100KHz both phases and bottom left is the same, just showing the sum across the load.
Now for a few latest waveforms...
The top left is a 300KHz
square wave at 10Vp per phase. bottom left is a 400KHz sine wave, Ch 2 is inverted and added to Ch 1 so this is what you see across the speaker. Top right is 100KHz both phases and bottom left is the same, just showing the sum across the load.Attachments
This is a larger scale of the 20 Hz square wave. This is 20V peak on 8 Ohms. I nearly burned up the resistor I used to test this one, and with the 300KHz square wave, left it on too long and kinda burned my finger. 
Next I will test the 20 Hz square wave with a speaker and see what the reactive load throws into the mix.
Next I will test the 20 Hz square wave with a speaker and see what the reactive load throws into the mix.Attachments
It was too late yesterday to crank out up a speaker with low tones, but here is 18 Hz square wave driving a 4 Ohm sub, 10V peak. The time base is 10mS, so the shutter of the camera caught a double trace, But you can see the shape well. There is a little bit of overshoot, but man, that cone was movin'. It sounded like a badly muffled car engine. 
I think the amp would drive a lower impedance as is, with a more beefy power supply. The problem is there are only 4 TO-220 outputs. They are quite rugged for their size, but I doubt they could do much more than 100W RMS due to SOA, even with an oversized heat sink. However, I see no reason why I can't use larger devices. This way I won't have to match mosfet transistors. It is nice to be able to just take one out of the package, slap it in the circuit and it works perfectly because the circuit operation is not dependent on specific parameters of a particular device. Since they are essentially in series, they require no matching.
That bit about not having to use a speaker relay is quite step. I would never have expected this circuit would have no DC output misbalance to the speaker when the power is switched on or off. A completely accidental discovery. I was planning to use a couple of UPC1237's or equivalent to drive a relay, which I would still have to find and experiment with.
This puts me much closer to making the PCB a reality.
I think the amp would drive a lower impedance as is, with a more beefy power supply. The problem is there are only 4 TO-220 outputs. They are quite rugged for their size, but I doubt they could do much more than 100W RMS due to SOA, even with an oversized heat sink. However, I see no reason why I can't use larger devices. This way I won't have to match mosfet transistors. It is nice to be able to just take one out of the package, slap it in the circuit and it works perfectly because the circuit operation is not dependent on specific parameters of a particular device. Since they are essentially in series, they require no matching.
That bit about not having to use a speaker relay is quite step. I would never have expected this circuit would have no DC output misbalance to the speaker when the power is switched on or off. A completely accidental discovery.
I was planning to use a couple of UPC1237's or equivalent to drive a relay, which I would still have to find and experiment with. This puts me much closer to making the PCB a reality.Attachments
Complicated? Naw, ....well maybe just a wee bit. The output's are for TO-220 OR TO-3P. The pads at B+/B- (25V) are for fuse holders. Was thinking of the 3557-2
TO-126 horizontal parts are the Vbe multipliers/EC.
The TO-126 vertical parts are the drivers, and the ones with 2 leads are Caddock resistors as output source resistors. It looks like it all runs together, but there are actually seperate circuit sections, somewhat symetrical.
My rules for the 201 size SMD resistors are, no more than 20V and no more than 20mW. There are many here but most are just a few mW. I derate the SOT-563 parts to 100mW although it is less in this circuit. I think there are two that are at 50mW, the rest are less. These things are great for CCS, among other things because they take very little space for two descrete transistors.
The whole circuit will use no more than 30mA, excluding the channels of the output mosfets. If it works as good as the Frankenstien circuit I have now, all the hours spent will have been worth it. Still many design checks to make yet...
EDIT:
TOP
The output's are for TO-220 OR TO-3P. The pads at B+/B- (25V) are for fuse holders. Was thinking of the 3557-2TO-126 horizontal parts are the Vbe multipliers/EC.
The TO-126 vertical parts are the drivers, and the ones with 2 leads are Caddock resistors as output source resistors. It looks like it all runs together, but there are actually seperate circuit sections, somewhat symetrical.
My rules for the 201 size SMD resistors are, no more than 20V and no more than 20mW. There are many here but most are just a few mW. I derate the SOT-563 parts to 100mW although it is less in this circuit. I think there are two that are at 50mW, the rest are less. These things are great for CCS, among other things because they take very little space for two descrete transistors.
The whole circuit will use no more than 30mA, excluding the channels of the output mosfets. If it works as good as the Frankenstien circuit I have now, all the hours spent will have been worth it. Still many design checks to make yet...EDIT:
TOP
Attachments
Triangle and clipping
Here are some triangle waveforms along with hard and soft clipping on the load at 50KHz. At hard clipping, the output stage(s) actually decrease bias into class B operation, so there is some crossover components, but it has to be continously driven into hard clipping for this to happen. In spite of the complexity, this seems to be the most stable amplifier I've made yet.
Here are some triangle waveforms along with hard and soft clipping on the load at 50KHz. At hard clipping, the output stage(s) actually decrease bias into class B operation, so there is some crossover components, but it has to be continously driven into hard clipping for this to happen. In spite of the complexity, this seems to be the most stable amplifier I've made yet.
Attachments
Hi Carlos
You seem to constitute the interest of this 'non-global feedback' circuit array. I think I'm getting close with the output stage circuit. Here is a sketch of the output stage section of circuitry so far. Q1 & Q2 is KSC2690 and Q3 & Q4 is KSA1220. 4) 1000uf 25V decoupling caps on the output drains are left out. All the devices are from Fairchild; up to this point MHO is they do make quality parts, both in the power and small signal genres.
The VAS stage is a balanced complementary constant power differential bridge,, and DC bias to the bases of each pre-driver is about +/-11V. The VAS has around 24V of swing possible beyond this DC bias.
Points A and B are the feedback points for the DC servo circuit. There is no AC signal global feedback, but there must be a DC feedback loop because after the first input caps to the first stage(s), the entire circuit is direct coupled all the way to the output, which requires DC bias control for each phase.
In case you didn't notice, this is an adaptation of Mr. Cordell's Mosfet output stage, just X 2, based on Dr. Hawksford's Darlington EC circuit. Works very well, thanks Bob! Total bias of each drive circuit is about 7mA. The outputs' bias is about 250mA at idle, but will decrease as the transistors get hotter (hot to the touch)down to 100mA. The reason the gate resistors aren't the same is because the output fet's aren't really complementary, but you got to love a component forgiving circuit when the budget for development is slim.
I put the circuit in the freezer for a while, and then hooked it up to measure bias. It started out at nearly an ampere, but quickly decreased back to about 250mA as the heat sink warmed. IMO, this behavior leads to a more thermally stable design.
It is an election day here in the US
and nobody represents me, there is a big storm coming with high winds and possible tornados leaving nothing else to do but
and watch the nature show!!
You seem to constitute the interest of this 'non-global feedback' circuit array.
I think I'm getting close with the output stage circuit. Here is a sketch of the output stage section of circuitry so far. Q1 & Q2 is KSC2690 and Q3 & Q4 is KSA1220. 4) 1000uf 25V decoupling caps on the output drains are left out. All the devices are from Fairchild; up to this point MHO is they do make quality parts, both in the power and small signal genres. The VAS stage is a balanced complementary constant power differential bridge,
, and DC bias to the bases of each pre-driver is about +/-11V. The VAS has around 24V of swing possible beyond this DC bias. Points A and B are the feedback points for the DC servo circuit. There is no AC signal global feedback, but there must be a DC feedback loop because after the first input caps to the first stage(s), the entire circuit is direct coupled all the way to the output, which requires DC bias control for each phase.
In case you didn't notice, this is an adaptation of Mr. Cordell's Mosfet output stage, just X 2, based on Dr. Hawksford's Darlington EC circuit. Works very well, thanks Bob!
Total bias of each drive circuit is about 7mA. The outputs' bias is about 250mA at idle, but will decrease as the transistors get hotter (hot to the touch)down to 100mA. The reason the gate resistors aren't the same is because the output fet's aren't really complementary, but you got to love a component forgiving circuit when the budget for development is slim. I put the circuit in the freezer for a while, and then hooked it up to measure bias. It started out at nearly an ampere, but quickly decreased back to about 250mA as the heat sink warmed. IMO, this behavior leads to a more thermally stable design.
It is an election day here in the US
and nobody represents me, there is a big storm coming with high winds and possible tornados leaving nothing else to do but and watch the nature show!! Attachments
Hi all
OK... I've finally decided on an output stage for my amp, something similar to this. Of course there will be two of these circuits driven at 180 degrees to bridge across the speaker. I have the top and bottom CCS referenced to the second cascode instead of the first because the first one may saturate if overdriven so the second will continue to function as a current source. In the Frankenstein prototype I'm listening to right now(sounds sweet and musical with lots of detail and sound stage) the first pre-driver and driver are not cascode, but I don't see where there is a problem as to why it wouldn't work...assuming I construct it properly.
This circuit really motivates me. I was listening to some classical music earlier and I swear I heard someone on stage clear the spit from their horn during an orchestral rest....
OK... I've finally decided on an output stage for my amp, something similar to this. Of course there will be two of these circuits driven at 180 degrees to bridge across the speaker. I have the top and bottom CCS referenced to the second cascode instead of the first because the first one may saturate if overdriven so the second will continue to function as a current source. In the Frankenstein prototype I'm listening to right now(sounds sweet and musical with lots of detail and sound stage) the first pre-driver and driver are not cascode, but I don't see where there is a problem as to why it wouldn't work...assuming I construct it properly.
This circuit really motivates me. I was listening to some classical music earlier and I swear I heard someone on stage clear the spit from their horn during an orchestral rest....
Attachments
This is how far I am on the new PCB since I decided to start over due to the fact I didn't like the layout. I still have yet to include the second DC servo, the +40V reg, and the -40V reg circuits. More work to due yet. It will take me days just to re-label the components to correspond with my master schematic drawing, but it will make it easier to construct. Always planning for the future...
Attachments
Hi
Almost finished...with the silk screen documentation and the first of many reviews. I found and corrected a couple of mistakes, progress is being made. I still have 1 circuit section to finish labeling parts and I'm already up to R169. I like this layout a whole lot better. I managed to get the 3 jumpers that selects balanced OR single end input together as 1 jumper, in between the input caps so it actually looks more symmetrical. The circuit itself actually is symmetrical, in more than one sense. I know it seems like using such tiny parts is ludicrous with such a complex design, but the self biasing circuit I discovered tends to limit current as opposed to over bias and smoke. I have yet to burn an SMD part with the prototyping since I started using this circuit idea. If the circuit tries to over bias, the CCS's drop the voltage out, never allowing the current to be more than the CCS. It won't function right like this, but at least nothing burns. This is the ONLY reason I was so successful with the Frankenstein construction. There was some "trial and error" used when finding the correct combination of resistor values and if the "error" had resulted in smoke, I would have never got it to work. All 201 size resistors have less than 20V, and less than 20mW, so I don't foresee any problem. Might need a microscope though....
At a glance, it looks like one huge circuit blob, but it is actually 8 completely separate discrete amplifiers, each one connected to another by only a few circuit paths, so I will assemble it just like I did the Frankenstein proto version, one sub-circuit at a time.
POS 40V reg
NEG 40V reg
2 DC servos
1 gain/phase spitting stage
1 VAS
2 power stages
The +/-50V powers everything except the output channels and speaker with about 30mA, a whopping whole 3 Watts. It can actually be +/-45V to +/-55V, with no difference in operation. It seems overall to be a fairly forgiving circuit. I wouldn't think there is a problem using 16 mil traces /w 1 oz copper as there is only a couple of mA flowing through it and most are uA current levels.
I am thinking of adding a couple of electrolytic decoupling caps for this voltage source but would there be a significant advantage by having these before the regulator circuit? or after? But if I put it after the reg circuits, I will have an electrolytic driving the circuit which defeats the purpose of using amplified polypropylene caps.
On a good note, I cranked up the proto circuit today to full power with the undersized heat sink to do some torture testing. The outputs got so hot the plastic case burned my finger a bit, and they still didn't explode. Bias dropped a bit, but I think I will go for a higher bias anyway, maybe 300-400mA. Right now it is 150mA, and dropped down to about 50mA when it was very hot. I was studying the overall logic of the amp and it seems that if an output were to short, the DC servos would force the other (bridged) side to the same potential as the short, which should result in zero potential across speaker. The source resistors on that side might burn, but the speaker should be safe. I haven't worked up the nerve to deliberately short out an output device to test that theory yet though. .........
Almost finished...with the silk screen documentation and the first of many reviews. I found and corrected a couple of mistakes, progress is being made. I still have 1 circuit section to finish labeling parts and I'm already up to R169.
I like this layout a whole lot better. I managed to get the 3 jumpers that selects balanced OR single end input together as 1 jumper, in between the input caps so it actually looks more symmetrical. The circuit itself actually is symmetrical, in more than one sense. I know it seems like using such tiny parts is ludicrous with such a complex design, but the self biasing circuit I discovered tends to limit current as opposed to over bias and smoke. I have yet to burn an SMD part with the prototyping since I started using this circuit idea. If the circuit tries to over bias, the CCS's drop the voltage out, never allowing the current to be more than the CCS. It won't function right like this, but at least nothing burns. This is the ONLY reason I was so successful with the Frankenstein construction. There was some "trial and error" used when finding the correct combination of resistor values and if the "error" had resulted in smoke, I would have never got it to work. All 201 size resistors have less than 20V, and less than 20mW, so I don't foresee any problem. Might need a microscope though....At a glance, it looks like one huge circuit blob, but it is actually 8 completely separate discrete amplifiers, each one connected to another by only a few circuit paths, so I will assemble it just like I did the Frankenstein proto version, one sub-circuit at a time.
POS 40V reg
NEG 40V reg
2 DC servos
1 gain/phase spitting stage
1 VAS
2 power stages
The +/-50V powers everything except the output channels and speaker with about 30mA, a whopping whole 3 Watts.
It can actually be +/-45V to +/-55V, with no difference in operation. It seems overall to be a fairly forgiving circuit. I wouldn't think there is a problem using 16 mil traces /w 1 oz copper as there is only a couple of mA flowing through it and most are uA current levels. I am thinking of adding a couple of electrolytic decoupling caps for this voltage source but would there be a significant advantage by having these before the regulator circuit? or after? But if I put it after the reg circuits, I will have an electrolytic driving the circuit which defeats the purpose of using amplified polypropylene caps.
On a good note, I cranked up the proto circuit today to full power with the undersized heat sink to do some torture testing. The outputs got so hot the plastic case burned my finger a bit, and they still didn't explode.
Bias dropped a bit, but I think I will go for a higher bias anyway, maybe 300-400mA. Right now it is 150mA, and dropped down to about 50mA when it was very hot. I was studying the overall logic of the amp and it seems that if an output were to short, the DC servos would force the other (bridged) side to the same potential as the short, which should result in zero potential across speaker. The source resistors on that side might burn, but the speaker should be safe. I haven't worked up the nerve to deliberately short out an output device to test that theory yet though. .........
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.