There's a really great amplifier, designed by Bob Cordell.
It's a FET input and MOSFET output.
a description can be found at:
http://www.cordellaudio.com/papers/mosfet_with_error_correction.shtml
I'd like to start a discussion of updating and building this amp.
It's only 50 watts as originally designed, but it should be easily modified to over 100watts just
by raising the supply voltage to 50 volts on the MOSFET outputs and maybe paralleling 2 of them.
(+raising the input supply to 65volts)
I'm not posting this merely to share my own adventures, but rather (i hope) to get a large participation of experiences and ideas.
It's a FET input and MOSFET output.
a description can be found at:
http://www.cordellaudio.com/papers/mosfet_with_error_correction.shtml
I'd like to start a discussion of updating and building this amp.
It's only 50 watts as originally designed, but it should be easily modified to over 100watts just
by raising the supply voltage to 50 volts on the MOSFET outputs and maybe paralleling 2 of them.
(+raising the input supply to 65volts)
I'm not posting this merely to share my own adventures, but rather (i hope) to get a large participation of experiences and ideas.
myhrrhleine said:
Bob's Hawkford EC output stage, modified for 100W/8ohm and 200W/4ohm, lateral MOSFETs, trimmed to 80ppm:
http://www.synaesthesia.ca/OP-stage.html
PCB Gerber files are also available (click the link).
Listening test (with the PGP front end):
http://www.synaesthesia.ca/listening.html
Hi
I find it difficult to get a perfectly stable bias current between start up and full temperature. With 200mA bias, I was able to get the bias fairly stable when the HS temp warmed slightly and for when it is hot, but the initial bias is a bit high. I don't think this is a problem if it is not too excessive, but others may have other opinions.
I didn't dwell too much on the issue. I think Bob mentioned somewhere that he only placed one of the EC transistors on the heat sink and had better tracking results. The challenge is to get the EC transistors off the heat sink. My thoughts were to use a small signal SOT-23 device and mount it on the PCB inderneth the output transistor pins, so as it contacts the drain pin right were it enters the component. I've also contemplated using a daughter board for the EC amplifier.
As for the operation of the output stage, it has great potential to get real results from fairly cheap components.
I find it difficult to get a perfectly stable bias current between start up and full temperature. With 200mA bias, I was able to get the bias fairly stable when the HS temp warmed slightly and for when it is hot, but the initial bias is a bit high. I don't think this is a problem if it is not too excessive, but others may have other opinions.
I didn't dwell too much on the issue. I think Bob mentioned somewhere that he only placed one of the EC transistors on the heat sink and had better tracking results. The challenge is to get the EC transistors off the heat sink. My thoughts were to use a small signal SOT-23 device and mount it on the PCB inderneth the output transistor pins, so as it contacts the drain pin right were it enters the component. I've also contemplated using a daughter board for the EC amplifier. As for the operation of the output stage, it has great potential to get real results from fairly cheap components.
CBS240 said:Hi
I find it difficult to get a perfectly stable bias current between start up and full temperature. With 200mA bias, I was able to get the bias fairly stable when the HS temp warmed slightly and for when it is hot, but the initial bias is a bit high. I don't think this is a problem if it is not too excessive, but others may have other opinions.
As for the operation of the output stage, it has great potential to get real results from fairly cheap components.
Hi CBS240,
When you were having difficulty with stabilizing the bias, was that without either of the EC transistors on the heat sink?
There's a graph of bias vs time in the original JAES paper that compares bias stability with that of a BJT amplifier, and the MOSFET one looks quite good. However, different implementations may need different degrees or types of temperature compensation. I think I biased the original 50W amplifier at 150 mA with 35-volt main rails.
I recognize that it would be very nice to not have to have one of the EC transistors on the heat sink, since they and their part of the circuit want to be optimized for best high-frequency performance. In my original design, I think I used something like an MPSU07 for the EC transistor that went on the heat sink because it had a convenient mounting tab. Obviously, this is not the fastest transistor in the world and the need to have some wires to it doesn't help the HF either. I've often pondered ways to get the EC transistor off the heat sink, but have never bothered with it.
Cheers,
Bob
Hi Bob
I have experimented with both transistors on the heatsink, then removing just one from the heatsink, to using a Vbe multiplier circuit in series with the EC amplifier. Even with one transistor on the heatsink, it would still track with a slight neg Tc, which is better than a pos Tc.
Increasing the emitter resistors on the error amp helped the tracking by reducing the Tc of the error amp transistors, but it also reduces the gain of the error amplifier and that would lead to a pointless outcome. To me, in my 'home-made' amp, I don't really care much if it over compensates a bit, it doesn’t decrease that much so I just set the 'cold' bias a little high, but it’s not perfect.
Pondering…..If you mount the outputs horizontal to the PCB so that the transistor pins bend 90 degrees into the PCB, then left the drain lead a bit longer than the others so that under this pin you could mount a SOT-23 (EC amplifier) in contact with it using a solder pad that would allow it to 'slide' from the case of the output transistor to a few mm away so as to use the drain pin lead as a pot in the thermal model instead of a fixed resistor, assuming that it over-compensates when mounted closest to the output transistor. Of course you want to keep the leads as short as possible to impede parasitics, but it might be worth the trade off on the drain pin to be able to easily tune in the thermal bias stability by simply moving the position of a couple of transistors. The question is, could it be done neatly, with all the other needed circuit components on less than 3 layers.
I have experimented with both transistors on the heatsink, then removing just one from the heatsink, to using a Vbe multiplier circuit in series with the EC amplifier. Even with one transistor on the heatsink, it would still track with a slight neg Tc, which is better than a pos Tc.
Increasing the emitter resistors on the error amp helped the tracking by reducing the Tc of the error amp transistors, but it also reduces the gain of the error amplifier and that would lead to a pointless outcome. To me, in my 'home-made' amp, I don't really care much if it over compensates a bit, it doesn’t decrease that much so I just set the 'cold' bias a little high, but it’s not perfect.Pondering…..If you mount the outputs horizontal to the PCB so that the transistor pins bend 90 degrees into the PCB, then left the drain lead a bit longer than the others so that under this pin you could mount a SOT-23 (EC amplifier) in contact with it using a solder pad that would allow it to 'slide' from the case of the output transistor to a few mm away so as to use the drain pin lead as a pot in the thermal model instead of a fixed resistor, assuming that it over-compensates when mounted closest to the output transistor. Of course you want to keep the leads as short as possible to impede parasitics, but it might be worth the trade off on the drain pin to be able to easily tune in the thermal bias stability by simply moving the position of a couple of transistors. The question is, could it be done neatly, with all the other needed circuit components on less than 3 layers.
CBS240 said:Hi Bob
I have experimented with both transistors on the heatsink, then removing just one from the heatsink, to using a Vbe multiplier circuit in series with the EC amplifier. Even with one transistor on the heatsink, it would still track with a slight neg Tc, which is better than a pos Tc.
Pondering…..If you mount the outputs horizontal to the PCB so that the transistor pins bend 90 degrees into the PCB, then left the drain lead a bit longer than the others so that under this pin you could mount a SOT-23 (EC amplifier) in contact with it using a solder pad that would allow it to 'slide' from the case of the output transistor to a few mm away so as to use the drain pin lead as a pot in the thermal model instead of a fixed resistor, assuming that it over-compensates when mounted closest to the output transistor. Of course you want to keep the leads as short as possible to impede parasitics, but it might be worth the trade off on the drain pin to be able to easily tune in the thermal bias stability by simply moving the position of a couple of transistors. The question is, could it be done neatly, with all the other needed circuit components on less than 3 layers.
This is an interesting suggestion, although mechanically it would be a bit non-standard. Maybe what we need is some kind of opto-isolated thermal current mirror or conveyor
.I have not tried it, but putting some kind of a thermal temperature coefficient in the fixed main bias spreader (the 22V spreader on the VAS output) might also work, since there is some effect it has on the final spread. I just haven't evaluated how much that effect is.
Cheers,
Bob
Bob Cordell said:
This is an interesting suggestion, although mechanically it would be a bit non-standard. Maybe what we need is some kind of opto-isolated thermal current mirror or conveyor
I have not tried it, but putting some kind of a thermal temperature coefficient in the fixed main bias spreader (the 22V spreader on the VAS output) might also work, since there is some effect it has on the final spread. I just haven't evaluated how much that effect is.
Cheers,
Bob
If you mean something like a Vbe multiplier in 'series' with the EC network, as the bias voltage spread source would see assuming it is constant, I haven't completely given up on that idea, yet I think it will be difficult to make perfect. I find that although I can get the Vbe multiplier to track the temperature of the outputs fairly well by adjusting its gain/emitter resistor, the temperature of the error correction transistors is relevant. In the circuit I was playing with, the bias starts out high due to this reason but then settles and tracks fairly close the set point because there are components on the opposing side of the PCB that supply a small bit of heat to the EC. If you create extra airflow on the EC transistors, the bias reacts to it. Maybe using a heater resistor in contact with the error devices to keep them at a constant temperature just above ambient would help, and could be a point of fine tune adjustment in the circuit with the seperate Vbe multiplier. Opto-coupling is an interesting thought, but how it correlates with the temperature of the error devices would be relevant. Perhaps some combination of these ideas would solve the problem but then, how to keep it simple?
I still like the OPS circuit because it is so easily expandable.
If the drive circuit is current limited, an overload pretty much blows the output hexfets (fuse included of course), so all you have to do is replace a $1 hexfet transistor....or use a bigger one....or use multiple hexfets and compensate accordingly.
m2003br said:
Hi m2003br
I had a look there, http://www.cordellaudio.com/papers/...orrection.shtml , but I do not seem to be able to find it. Is it on a particular page in the pdf that I simply am not recognising?
Bob Cordell said:
This is an interesting suggestion, although mechanically it would be a bit non-standard. Maybe what we need is some kind of opto-isolated thermal current mirror or conveyor
I have not tried it, but putting some kind of a thermal temperature coefficient in the fixed main bias spreader (the 22V spreader on the VAS output) might also work, since there is some effect it has on the final spread. I just haven't evaluated how much that effect is.
Cheers,
Bob
Hi! I´m also experimenting with your EC output, and also had some problems stabilizing the bias. I did actually use a 15 volt zener and a multiplier (mounted onto the heat sink) in series in place of the original 22 volt zener, and it does track quite well now. I havent placed the EC transistors on the heat sink.
One question to you Bob, have you measured any major performance difference between different transistors in the EC stage? I presume using fast ones will improve the correction at high frequencies, but would appreciate any suggestions.
Rgs Rikard
Just tried some things: A simple multiplier in series with the zener does the trick! You can determine the rate of temp coeff by altering the voltage across the zener in relation to the multiplier. I got good results from a 15 volt Zener and 5-7 volts across the multiplier. Seems stable enough.
By the way, I´m using IRFP240/9240s for the output.
By the way, I´m using IRFP240/9240s for the output.
KLe said:
Hi m2003br
I had a look there, http://www.cordellaudio.com/papers/...orrection.shtml , but I do not seem to be able to find it. Is it on a particular page in the pdf that I simply am not recognising?
Here is the paper, the output circuit is on page 10, the other part is on page 8.
Rikard Nilsson said:
Hi Rikard,
No, but I do believe that we can do much better than the garden variety ones I used. Transistors with ft on the order of 1 Ghz would likely yield an improvement if their beta is decent. We obviously do not need high-voltage transistors for either the EC stage or the EFs following it.
Cheers,
Bob
Thanks for your reply Bob! The prototype I built use MJE340/350 as I had some lying around, but it doesn´t seem to work 100%, I have tried using trimpots but still cannot find a "dip" in the distorsion. Perhaps these trannys are simply not good enough, or I have missed something. As I understand I should get a well defined "error correction balance" tuning the pots?
Rikard Nilsson said:Thanks for your reply Bob! The prototype I built use MJE340/350 as I had some lying around, but it doesn´t seem to work 100%, I have tried using trimpots but still cannot find a "dip" in the distorsion. Perhaps these trannys are simply not good enough, or I have missed something. As I understand I should get a well defined "error correction balance" tuning the pots?
If things are set up properly, yes, you should get a well-defined dip in distortion as the EC trimpot is varied. However, if the setup is in error to one side, you might just see distortion moving up or down monotonically in one direction. If that were the case, and the rest of the toplogy were working right, you should make things so that the fixed resistances in the circuit get you near the low-distortion end of the trimpot setting when the trimpot is at the other extreme. In other words, you are "walking" the fixed component values in the direction of lower distortion that will bring the adjustment range of the trimpot to where it can find the minima.
Cheers,
Bob