vuthanh6 said:
Hi Thanh,
With rails of 53v the heatsink will be very warm but nowhere near warm enough to worry. I run a similar heatsink to yours with rails of 73v and my amp has been running since Feb 2005 without any problems at all.
Nice photos of your amp, it looks very professional. How does it sound. Are you happy with it?
Cheers
Quasi
Re[03]: First monoblock is working
analog_guy,
Excellent work in the construction of your two monoblocks and the case.
Is it true, from what I read at verious times, that when using laterial MOSFETS one really does not need bias tracking (such as what T8 provides in quasi's design for vertical MOSFETS) of the output stage?
Am I correct in observing the heatsinks for the output devices are a solid block of aluminum?
I, and I think quasi will be as well, interested to know if the +/- 78V that supplies the "gain stage" is to just the LTP or LTP, VAS and driver stages? I assume because you stated "+/- 78V" supply that both the negative and positive rails of the "gain stage" are boosted the +/-12V? The reason I am interested, and I suspect quasi will be interested as well, is because quasi put much effort into the design initially to use a split PSU configuration (as it appears you have done) that your testing has indicated sounds great and is stable.
My interest in a split PSU configuration is because I want to use rails in the order of +/- 30-35V for the rails that supply the output devices. It is easy for me to add an additional winding to some toroids I have to provide a higher voltage to the LTP, VAS and driver stage to reduce output stage losses (about 12V-17V) when using a single rail voltage to supply the entire amplifier module. As I am sure you are aware, when using rail voltages as I would like to use the 12V-17V loss is a very high loss on a percentage/dB basis in the output power the amplifier module can deliver.
That said, I am sure there are many other builders of quasi's NMOS design that would like to know how you accomplished a successfull implementation of a split PSU of quasi's design.
Based on quasi's efforts and findings using +/- 60V for the output stage and +/- 70V for the LTP, VAS and driver stage verses your results with the variant quasi design you have completed can you share why your split rail version appears not to have the clipping issues quasi experienced:
Regards,
John L. Males
Willowdale, Ontario
Canada
29 December 2007 (09:30) - 10:46
Official Quasi Thread Researcher
analog_guy said:
analog_guy,
Excellent work in the construction of your two monoblocks and the case.
Is it true, from what I read at verious times, that when using laterial MOSFETS one really does not need bias tracking (such as what T8 provides in quasi's design for vertical MOSFETS) of the output stage?
Am I correct in observing the heatsinks for the output devices are a solid block of aluminum?
I, and I think quasi will be as well, interested to know if the +/- 78V that supplies the "gain stage" is to just the LTP or LTP, VAS and driver stages? I assume because you stated "+/- 78V" supply that both the negative and positive rails of the "gain stage" are boosted the +/-12V? The reason I am interested, and I suspect quasi will be interested as well, is because quasi put much effort into the design initially to use a split PSU configuration (as it appears you have done) that your testing has indicated sounds great and is stable.
My interest in a split PSU configuration is because I want to use rails in the order of +/- 30-35V for the rails that supply the output devices. It is easy for me to add an additional winding to some toroids I have to provide a higher voltage to the LTP, VAS and driver stage to reduce output stage losses (about 12V-17V) when using a single rail voltage to supply the entire amplifier module. As I am sure you are aware, when using rail voltages as I would like to use the 12V-17V loss is a very high loss on a percentage/dB basis in the output power the amplifier module can deliver.
That said, I am sure there are many other builders of quasi's NMOS design that would like to know how you accomplished a successfull implementation of a split PSU of quasi's design.
Based on quasi's efforts and findings using +/- 60V for the output stage and +/- 70V for the LTP, VAS and driver stage verses your results with the variant quasi design you have completed can you share why your split rail version appears not to have the clipping issues quasi experienced:
and the clipping isn't as clean.Regards,
John L. Males
Willowdale, Ontario
Canada
29 December 2007 (09:30) - 10:46
Official Quasi Thread Researcher
Reply to John Males
John,
you ask a number of interesting questions, some of which I also encountered during the development process and for which I also had to find solutions.
Regarding thermal tracking, I did notice that the bias current through the MOSFETS did vary with temperature, but with a strongly negative coefficient. For example, I am using a P Ch and N-ch pair of IRF devices as a follower and have seen up to a 30% reduction in Ibias from cold to warm up. That said, I did not observe any degradation in performance over the range of Ibias that was encountered. For this reason I did not apply any Ibias correction to the design.
The heatsinks are extrusions with the fins running parallel to the floor. The photo does not show this well.
The power supply picture is somewhat complicated. I chose to boost the +/- 65V supply to a higher value for a couple of reasons. The first is that it guarantees that there is sufficient voltage available to drive the output stage. These MOSFETS typically require 3-5V Vgs to achieve max output, and this voltage must deducted from the available VAS rail voltage. There is an additional voltage drop caused by the cascode topology in the VAS stage.
The second reason is particular to the EC design topology. I drive the EC amp from a MOSFET follower on the VAS satge, and by maintaining a minimum of ~12V Vds, the follower devices' capacitance is approximately constant, something that is not the case as Vds goes below 10V.
The +/- 78 supplies the VAS and the EC follower stages only. The LTP rails are regulated down to +/- 55V.
Implementing a split PS does require careful attention to power and power return paths, In this case I built a separate power distribution PCB that sits atop the large electrolytics. The boost and EC supplies are built with isolated DC-DC converters operating off the +/- 65V rails. All the PC boards are 2-layers, which allows a good ground plane. All grounds are returned to this PC board, and all power voltages are decoupled to the local PCB ground plane.
Regarding clipping: I decided to build a simple diode clamp outside the feedback loop at the amp's input. The two diodes connect to a +/- 4.5V zener derived supply. This approach solves two problems. It guarantees symmetrical clipping independent of the amp's internal topology. It also guarantees that when clipping does occur, there is no recovery time penalty as might be the case if clipping occurred inside the FB loop. For amps with a large loop gain, clipping forces the amp's internal voltages far away from their quiescent points. When clipping is removed it can take a substantial time for the amp to re-establish the quiescent voltage/currents, and during this time the amp may produce increased distortion.
John,
you ask a number of interesting questions, some of which I also encountered during the development process and for which I also had to find solutions.
Regarding thermal tracking, I did notice that the bias current through the MOSFETS did vary with temperature, but with a strongly negative coefficient. For example, I am using a P Ch and N-ch pair of IRF devices as a follower and have seen up to a 30% reduction in Ibias from cold to warm up. That said, I did not observe any degradation in performance over the range of Ibias that was encountered. For this reason I did not apply any Ibias correction to the design.
The heatsinks are extrusions with the fins running parallel to the floor. The photo does not show this well.
The power supply picture is somewhat complicated. I chose to boost the +/- 65V supply to a higher value for a couple of reasons. The first is that it guarantees that there is sufficient voltage available to drive the output stage. These MOSFETS typically require 3-5V Vgs to achieve max output, and this voltage must deducted from the available VAS rail voltage. There is an additional voltage drop caused by the cascode topology in the VAS stage.
The second reason is particular to the EC design topology. I drive the EC amp from a MOSFET follower on the VAS satge, and by maintaining a minimum of ~12V Vds, the follower devices' capacitance is approximately constant, something that is not the case as Vds goes below 10V.
The +/- 78 supplies the VAS and the EC follower stages only. The LTP rails are regulated down to +/- 55V.
Implementing a split PS does require careful attention to power and power return paths, In this case I built a separate power distribution PCB that sits atop the large electrolytics. The boost and EC supplies are built with isolated DC-DC converters operating off the +/- 65V rails. All the PC boards are 2-layers, which allows a good ground plane. All grounds are returned to this PC board, and all power voltages are decoupled to the local PCB ground plane.
Regarding clipping: I decided to build a simple diode clamp outside the feedback loop at the amp's input. The two diodes connect to a +/- 4.5V zener derived supply. This approach solves two problems. It guarantees symmetrical clipping independent of the amp's internal topology. It also guarantees that when clipping does occur, there is no recovery time penalty as might be the case if clipping occurred inside the FB loop. For amps with a large loop gain, clipping forces the amp's internal voltages far away from their quiescent points. When clipping is removed it can take a substantial time for the amp to re-establish the quiescent voltage/currents, and during this time the amp may produce increased distortion.
Re: Reply to John Males
Hello analog_guy,
Thanks for your reply. Very informative, though some bits are above my head, but that is ok.
An special reason you placed the heatsink fins facing down?
Correct me if I am wrong, but you modified the quasi NMOS design in part from quasi-complementary to a complementary design? Other than the PSU design and voltages is sounds like you removed the T8 bias network, added a diode network to the LTP stage, changed to a cascode VAS stage, used a MOSFET follwer to drive the EC amp, a EC amp, and N and P channel follower stage. Sounds like we have a new design? Is this the "Uncle of Quasi" or the "Aunt of Quasi"?? 😉
You clearly have put much effort, thought and experience into your design based on quasi's design. Would be fair to say because you changed the design to a complementary design that may be the reason you will not experience the asymmetrical clipping quasi could not resolve with the quasi-complementary output stage?
Your PSU voltage splitting is interesting. It is the first time I have heard of such a PSU topology approach. Due to quasi's findings I was thinking maybe there might be some benefit to running the LTP/VAS at a higher rail voltage and the rest at the output stage voltage. I had hoped I could do a simulation of some different approaches, but sadly I cannot find an ECAD/SPICE program that I can use under Linux to evaluate my ideas. I was not looking for perfection in the ECAD/SPICE program, just a reasonable approximation to help determine if any possible solutions exist. When I have more time I will have to design a unique prototype PCB so I can explore the different ideas I had wished to try out in ECAD/SPICE. I am recovering from a fractured foot so many things are on hold and as a consequence many things will have time setbacks until I am recovered and have caught up on many things not done during the injury.
Can you enlighten us why you chose to implement a complementary variant of the quasi NMOS? Personally I have always liked the N-Channel designs, but again that is a personal preference. I ask as I like to understand other's preferences and their reasons for their preferences. Just know when I ask questions it is alwys in context of understanding, not to suggest your preference/opinion is invalid.
I would be interested in seeing a schematics of the amplifier design you derived from the quasi NMOS design and the PSU design.
Again thanks so much for your informative reply. At the end of the day what is important is you are happy with your effort and design choices. For sure you will be happy about your case design once it is anodized and the monoblocks are all back together.
Regards,
John L. Males
Willowdale, Ontario
Canada
29 December 2007 (18:00 -) 18:51
Official Quasi Thread Researcher
P.S. I hope and trust everyone has had or will have a good holiday season and that you will have a safe season. jlm
Hello analog_guy,
Thanks for your reply. Very informative, though some bits are above my head, but that is ok.
An special reason you placed the heatsink fins facing down?
Correct me if I am wrong, but you modified the quasi NMOS design in part from quasi-complementary to a complementary design? Other than the PSU design and voltages is sounds like you removed the T8 bias network, added a diode network to the LTP stage, changed to a cascode VAS stage, used a MOSFET follwer to drive the EC amp, a EC amp, and N and P channel follower stage. Sounds like we have a new design? Is this the "Uncle of Quasi" or the "Aunt of Quasi"?? 😉
You clearly have put much effort, thought and experience into your design based on quasi's design. Would be fair to say because you changed the design to a complementary design that may be the reason you will not experience the asymmetrical clipping quasi could not resolve with the quasi-complementary output stage?
Your PSU voltage splitting is interesting. It is the first time I have heard of such a PSU topology approach. Due to quasi's findings I was thinking maybe there might be some benefit to running the LTP/VAS at a higher rail voltage and the rest at the output stage voltage. I had hoped I could do a simulation of some different approaches, but sadly I cannot find an ECAD/SPICE program that I can use under Linux to evaluate my ideas. I was not looking for perfection in the ECAD/SPICE program, just a reasonable approximation to help determine if any possible solutions exist. When I have more time I will have to design a unique prototype PCB so I can explore the different ideas I had wished to try out in ECAD/SPICE. I am recovering from a fractured foot so many things are on hold and as a consequence many things will have time setbacks until I am recovered and have caught up on many things not done during the injury.
Can you enlighten us why you chose to implement a complementary variant of the quasi NMOS? Personally I have always liked the N-Channel designs, but again that is a personal preference. I ask as I like to understand other's preferences and their reasons for their preferences. Just know when I ask questions it is alwys in context of understanding, not to suggest your preference/opinion is invalid.
I would be interested in seeing a schematics of the amplifier design you derived from the quasi NMOS design and the PSU design.
Again thanks so much for your informative reply. At the end of the day what is important is you are happy with your effort and design choices. For sure you will be happy about your case design once it is anodized and the monoblocks are all back together.
Regards,
John L. Males
Willowdale, Ontario
Canada
29 December 2007 (18:00 -) 18:51
Official Quasi Thread Researcher
P.S. I hope and trust everyone has had or will have a good holiday season and that you will have a safe season. jlm
Should we ask the moderators to extract these off topic posts and start a new thread for Analog?
Move Threads to Another Topic
Andrew,
That might be a good idea. I would leave it up to the moderator whether to include in an existing topic or start a new topic.
Jeff
AndrewT said:
Andrew,
That might be a good idea. I would leave it up to the moderator whether to include in an existing topic or start a new topic.
Jeff
NMOS200 running!
Finally, after a long long construction period, my NMOS200 amp modules are running as integrated amps (coupled to the OLD Leach line preamp). It's HOT inside the chassis, especially the KSE340/350, even with heatsinks as recommended on them. So, instead of passive cooling, I used 2 computer fans (series connected with dropping resistor to +50V rail) mounted on the top cover. Keeps things cooler inside. Added benefit is that the bias does not go down to less than 35mA (when hot) from around 65mA (cold), but stays at around 55 to 60mA. Still finishing off the top cover, maybe during the new year holidays. Happy New Year 2008 to all members!
Finally, after a long long construction period, my NMOS200 amp modules are running as integrated amps (coupled to the OLD Leach line preamp). It's HOT inside the chassis, especially the KSE340/350, even with heatsinks as recommended on them. So, instead of passive cooling, I used 2 computer fans (series connected with dropping resistor to +50V rail) mounted on the top cover. Keeps things cooler inside. Added benefit is that the bias does not go down to less than 35mA (when hot) from around 65mA (cold), but stays at around 55 to 60mA. Still finishing off the top cover, maybe during the new year holidays. Happy New Year 2008 to all members!
Hi guys, there is a problem with the amp. I switched on (with a light bulb in series with the primary winding of the transformer) and there were no fireworks. I checked the output offset with a DC meter and it was hovering around 5.5mV. Then I placed the DC voltmeter across the 100ohm resistor and turned the 200ohm trimmer. But the voltmeter shows 0V no matter what the position of the trimmer. Both channels have the same problem. Where should I start looking?
And wish all of you a happy 2008.
Vivek
And wish all of you a happy 2008.
Vivek
I also wish everyone a happy 2008, midnight (and thus the new year) already starts to propagate around the globe.
Keep up the good work, everybody!
Sebastian.
Keep up the good work, everybody!
Sebastian.
Seasons greetings to all of you
I finally got my NMOS 350 going. Came up with some problems. Even with a low bias current (15ma) the fets are getting very hot at medium to high music levels. I've tried it with both two pairs as well as one pair. My supply rails are +/- 65V dc. In fact at normal bias current levels the fets just blew at high music levels (I have now a collection of blown fets
)
Right now I'm running one pair at 15ma bias. Though the sound is simply brilliant (the bass goes real low and clean), the heating of the fets worries me. I am not driving it any higher.
If you recall, I'm using the NMOS 200 board layout for the NMOS 350 circuit. Could this have any bearing on my problem?
Regarding T8, is it supposed be thermally tracking the output pairs? If so would it be prudent to use a MJE340 instead? Also, in the NMOS 200 schematic there is a 0.1uf cap across the emitter-collector of T8 but not in the NMOS 350 schematic. I've included this in my layout. Could this have any bearing?
Hari
I finally got my NMOS 350 going. Came up with some problems. Even with a low bias current (15ma) the fets are getting very hot at medium to high music levels. I've tried it with both two pairs as well as one pair. My supply rails are +/- 65V dc. In fact at normal bias current levels the fets just blew at high music levels (I have now a collection of blown fets
)Right now I'm running one pair at 15ma bias. Though the sound is simply brilliant (the bass goes real low and clean), the heating of the fets worries me. I am not driving it any higher.
If you recall, I'm using the NMOS 200 board layout for the NMOS 350 circuit. Could this have any bearing on my problem?
Regarding T8, is it supposed be thermally tracking the output pairs? If so would it be prudent to use a MJE340 instead? Also, in the NMOS 200 schematic there is a 0.1uf cap across the emitter-collector of T8 but not in the NMOS 350 schematic. I've included this in my layout. Could this have any bearing?
Hari
Hi Hari,
Regarding your posting and questions I would say it is not normal for the MOSFETs to be running hot. I would share your concern that running a higher bias or higher signal levels will blow the MOSFETs. A few have chosen to run a higher bias for the NMOS350. That said the fact you cannot bias above 15ma and at 15ma bias cannot use a moderate signal level clearly confirms your problem.
What follows are some thoughts based on my limited knowledge and it is possible you have thought of and checked some o the suggestions I propose below. I am sure the experts will have some thoughts and comments to my suggestions below.
Have you checked using a lightblub or scope for any oscillations? I assume you have checked and rechecked your PCB and parts/soldering for any possible elements that might be related to the problem? It is also possible oscillations may be caused by the speaker cabling/speaker reactive loading or even as an antenna to some RF transmission you might not be keenly aware of. Have you tried using a resistive test load to see if you encounter the same warm/hot/bias limitations you currently have observed? Is it possible the source might be the input? Of course if osciillations is the problem one can have more than once source which would compound the problem.
Have you tried running your NMOS200 modules with a lower rail voltage just to see if there is any diference in the heat and bias issues?
It might be the PCB you designed is prone to oscillations. I cannot say if that would be the case or not. My hunch would be not as you did have the PCB screened for its design by the thread. Even so, it might be worth seeing if you built a NMSO200 using the quasi NMOS200 TO-247 layout if that would demonstrate any difference. Doing so might allow you to compare measurements between your PCB module and the quasi layout. Of course you will need to use MJE's and not BD's for the rail voltages you are using.
T9 and T10 will run warmer than most, if not all, other NMOS200 builds as you have a NMOS200 layout running as a NMOS300 in terms of rail voltages. I would suspect T9 and T10 may make it appear your MOSFETs are running warm to hot as a result. If you have a spare heatsink that would enable you to unscrew T9 and T10 from the main heatsink (bracket) it might be worth testing to see if having T9 and T10 seperated from the MOSFET heatsink just to see if there is any diference in the problem you are having assuming you find no oscillation issue using the lightblub or scope.
I assume you have checked that the BC546 T8 is thermally connected to the heatsink bracket? I would think if the hole was a tad too big this may add alot of thermal resistance which might be part of the reason for the warm/hot MOSFETs and bias issues. You can use a MJE340, BD139, BD137, or a few other TO-126 case types instead of the BC546 for T8. Some builders have used a TO-126 case for T8 to better ensure thermal connection to the MOSFET heatsink. In those cases the builders had no issue with choosing a TO-126 type T8 device. Using a TO-126 device for T8 certainly increases the confidence one has a good thermal connection to the MOSFET heatsink. If you only have MJE340's about in the TO-126 a MJE340 will be just fine. T8 does not have much power demand so a less power capable TO-126 device would make sense. Even a BD139 is well over the rating needed for T8, but most TO-126 devices will be. The MJE340s would be a last choice as they are one of the few devices you can use in the NMOS for the driver/pre-driver/VAS stages for the rail voltages you are using.
All the best for you and diyAudio members for 2008, especially those that have been so helpful to me in this thread. I do wish you are able to get your NMOS350 in a NMOS200 like layout to work just like a NMOS350. I am sure there are other builders that would like to use your PCB design.
Regards,
John L. Males
Willowdale, Ontario
Canada
01 January 2008 (09:55 -) 11:55
Official Quasi Thread Researcher
Regarding your posting and questions I would say it is not normal for the MOSFETs to be running hot. I would share your concern that running a higher bias or higher signal levels will blow the MOSFETs. A few have chosen to run a higher bias for the NMOS350. That said the fact you cannot bias above 15ma and at 15ma bias cannot use a moderate signal level clearly confirms your problem.
What follows are some thoughts based on my limited knowledge and it is possible you have thought of and checked some o the suggestions I propose below. I am sure the experts will have some thoughts and comments to my suggestions below.
Have you checked using a lightblub or scope for any oscillations? I assume you have checked and rechecked your PCB and parts/soldering for any possible elements that might be related to the problem? It is also possible oscillations may be caused by the speaker cabling/speaker reactive loading or even as an antenna to some RF transmission you might not be keenly aware of. Have you tried using a resistive test load to see if you encounter the same warm/hot/bias limitations you currently have observed? Is it possible the source might be the input? Of course if osciillations is the problem one can have more than once source which would compound the problem.
Have you tried running your NMOS200 modules with a lower rail voltage just to see if there is any diference in the heat and bias issues?
It might be the PCB you designed is prone to oscillations. I cannot say if that would be the case or not. My hunch would be not as you did have the PCB screened for its design by the thread. Even so, it might be worth seeing if you built a NMSO200 using the quasi NMOS200 TO-247 layout if that would demonstrate any difference. Doing so might allow you to compare measurements between your PCB module and the quasi layout. Of course you will need to use MJE's and not BD's for the rail voltages you are using.
T9 and T10 will run warmer than most, if not all, other NMOS200 builds as you have a NMOS200 layout running as a NMOS300 in terms of rail voltages. I would suspect T9 and T10 may make it appear your MOSFETs are running warm to hot as a result. If you have a spare heatsink that would enable you to unscrew T9 and T10 from the main heatsink (bracket) it might be worth testing to see if having T9 and T10 seperated from the MOSFET heatsink just to see if there is any diference in the problem you are having assuming you find no oscillation issue using the lightblub or scope.
I assume you have checked that the BC546 T8 is thermally connected to the heatsink bracket? I would think if the hole was a tad too big this may add alot of thermal resistance which might be part of the reason for the warm/hot MOSFETs and bias issues. You can use a MJE340, BD139, BD137, or a few other TO-126 case types instead of the BC546 for T8. Some builders have used a TO-126 case for T8 to better ensure thermal connection to the MOSFET heatsink. In those cases the builders had no issue with choosing a TO-126 type T8 device. Using a TO-126 device for T8 certainly increases the confidence one has a good thermal connection to the MOSFET heatsink. If you only have MJE340's about in the TO-126 a MJE340 will be just fine. T8 does not have much power demand so a less power capable TO-126 device would make sense. Even a BD139 is well over the rating needed for T8, but most TO-126 devices will be. The MJE340s would be a last choice as they are one of the few devices you can use in the NMOS for the driver/pre-driver/VAS stages for the rail voltages you are using.
All the best for you and diyAudio members for 2008, especially those that have been so helpful to me in this thread. I do wish you are able to get your NMOS350 in a NMOS200 like layout to work just like a NMOS350. I am sure there are other builders that would like to use your PCB design.
Regards,
John L. Males
Willowdale, Ontario
Canada
01 January 2008 (09:55 -) 11:55
Official Quasi Thread Researcher
Re[03]: Reply to John Males
Hi Jeff,
I actually almost completed my reply to you the later part of the afternoon of 30 December 2007, but the reply was lost before I was to send just a few minutes later, but the browser crashed and lost the reply contents. As I was already way behind on my day do to the pre-browser crash problems that had been ongoing and yet more browser problems I am not replying, hopefully posted, until now. Later I will recompile the browser from source so I have a binary I know matches my system libraries. The last time I recompiled the browser it too several hours, so it will be an overnight compile I will do.
Thanks for your very prompt reply. I have a some comments/questions as a result.
My comment of the heatsink facing down was because it appered there were no fins facing to the side, vertically or horizontially. Your comment about the pics you posted may not have shown the side of the heatsinks well in that regard makes perfect sense upon reflection.
I fully understand chosing surplus for heatsinks as opposed to buying new. When I can and it makes sense I use surplus. I have bought some excellent capacitors on surplus as well as some very good toroids. Both saved me lots of money, while still be excellent grade parts. I am going to make my heatsinks as other than one style of heatsink I cannot make, that is best choice, it is far less expensive and will allow me some flexibility in dimensions and thermal design. I have not been able to find even a few heatsinks to work with the NMOS350 design in surplus locally. I need to make at least 21 amps for a HT 3-way active crossover based system. Though I have read recently, posted some time ago, that a 4-way active is better for a number of reasons. I will have to research the difference as it is just not cost of amps, but a 4-way crossover active or not is far more complex to execute. There is little difference in building a 3 or 4-way active crossover.
Even though you did not use this thread as a point of reference and mistakenly posted to the thread I still think the design elements you used are of value to this thread. Sometimes a mistake happens to not be such mistake after all. At least in this case and in my opinion. If a moderator decides to pull your posing to another thread I sure hope they replace your posting in this thread with a link. there have already been a number of clearly off thread topic postings to this thread and other threads I have read and they were not pulled to a different thread. If you open a new thread or a moderator does, pulling these postings or not, I for one would like to know if you opened a new thread on your design.
Your insight about "quasi complementary designs might be more subject to asymmetrical clipping" and checking your symmetrical design without the diodes which confirmed the clipping was still symmetrical was very helpful.
Based on your comments for the TLP regulated and the are the same voltage as the output stae enlightens a previous idea I wanted to test. I will now do so at the same rail voltage and of course regulated. This may be the best compromise for the quasi complementary design of the quasi NMOS. I will likely then use a seperate bridge/filter capacitor for the driver stage.
Based on your PSU topology I would like to clarify your deisgn PSU in summary the LTP is a regulated +/-55, the VAS/Driver stage is +/-78V and the output stage is +/-65V? Assuming one has an output rail voltage of +/-X volts, how would you calculate what you would use for the LTP and VAS/Driver stage voltages as Y and Z respectively? I also have to assume one of more of these voltages factor into the amount of gain you design and how you determine the gain you set the amplifier to? I may be incorrectly assuming your add in a "safety margin" of the gain vs the PSU supply voltages? Then again maybe I should not assume 😉
My expereinces with LTspice were not great. LTspice seems to run well under the WINE emulator of Linux. A few problems exist, mostly related to selecting a part. The most important part and for reasons unknown the WINE emulator has some problems with this basic Windows function. The WIME emulator issues aside LTspice had some really basic problems. The first was when I entered the quasi NMOS design in LTspice from scratch and he models of parts not in LTspice. LTspice could not simulate or work with it at all. When I voiced that experience on this thread a gent posted his LTspice of the NMOS350. To that point I was never aware anyone had put quasi's NMOS350 into LTspice. His was the same as mine, except his was missing a few parts that were optional, some of the other values he used were different. Even so and with me adding in the few optional parts and normalizing the parts values LTspice could simulate the NMOS350. At least what I thought until I made a really gross value change to a key resistor that I am told should cause the TLP not to work. Well LTspice still simlated the design with the gross value change in a LTP resistor as if nothing had been changed! Not good as that resistor I wanted to fine tume for the rail rail voltages I wanted to test with. For reasons I will not understand LTspice seems not to understand the importance of the LTP resistor so I stopped using LTspice. Hence I am back to square zero in finding a ECAD/GUI based SPICE program that I can use and works using Linux.
I can understand why you feel and like a fully complementary amplifier design. For me the reasons for choosing a N-Channel design were it is hard for me to obtain complementary BiPolar ouput devices, notwithstanding the fact they are more expensive even for my modest power needs. Too many problems with counterfeit BiPolar output devices. I thouh I bought the genuine set of devices to build some BiPolar amps, but one side of the apir was very suspect. I do not have a RLC bridge to make a definitve conslusion. I seem to like the more robust nature of MOSFETs which also avoids the secondary breakdown voltage character of BiPolars. Thre are not as many complementary MSOFETs devices, let alone availability for me. There are alot of N-Channel MOSFETs I can purhcase locally. Also for right or wrong reasons, BiPolar or MOSFET I like the idea of the same output device on the + and - rail.
I am aware of the laterial 2SK/2SJ and ECX/ECN MOSFET devices. I was not aare of the BUZ901/906 being laterial devices. As much as I might be interested in using laterial MOSFETs, even at modest power handing abilities I need I just cannot afford these devices for the number of amplifiers I need to build. If I win the lotery well that would be a different story! lol lol lol I could afford and I still build my own amps and still build the quasi amps.
That said I was lucky about 8 months ago to come across some commerical receivers that have toroids. use the 2SC5200/2SA1943 output devices and 2SD669A/2SB649A drivers. What is so great was they all had a jumper to allow one to bypass the internal preamp, full speaker binding posts and at a price that beat all the brand names ignoring the china knock off devices they were using. These amps have excellent protection circuits and a inrush power on delay as well. Price was $80.00 CND! I knew there was a 6 channel HT version, but could not find one not the store carrying the stero version carry the HT version for reasons unknown.. Then I found a HT version and picked it up. Aside from the features I noted for the stereo it also had a dedicated 6 channel input which is great match for my Yahaha DSP-3000. this means I have a reference for how a Bipolar amp will sound, and once I build some quasi NMOSs I will be able to compare.
If I do get the time the best comparison would be to build the Brother of Quasi and then compare as this would be basically comparing the same amplifier, one with MOSFET outputs and one with BiPolar. Quasi even has a Class A in the wings of a similar design and uses BiPolar outputs. (Hint, maybe quasi will design a MOSFET output version of the Class A.)
Regards,
John L. Males
Willowdale, Ontario
Canada
01 January 2008 (12:00 - 12:30)
01 January 2008 (14:00 -) 16:17
01 January 2008 16:39 Typo corrections
Official Quasi Thread Researcher
Hi Jeff,
I actually almost completed my reply to you the later part of the afternoon of 30 December 2007, but the reply was lost before I was to send just a few minutes later, but the browser crashed and lost the reply contents. As I was already way behind on my day do to the pre-browser crash problems that had been ongoing and yet more browser problems I am not replying, hopefully posted, until now. Later I will recompile the browser from source so I have a binary I know matches my system libraries. The last time I recompiled the browser it too several hours, so it will be an overnight compile I will do.
Thanks for your very prompt reply. I have a some comments/questions as a result.
My comment of the heatsink facing down was because it appered there were no fins facing to the side, vertically or horizontially. Your comment about the pics you posted may not have shown the side of the heatsinks well in that regard makes perfect sense upon reflection.
I fully understand chosing surplus for heatsinks as opposed to buying new. When I can and it makes sense I use surplus. I have bought some excellent capacitors on surplus as well as some very good toroids. Both saved me lots of money, while still be excellent grade parts. I am going to make my heatsinks as other than one style of heatsink I cannot make, that is best choice, it is far less expensive and will allow me some flexibility in dimensions and thermal design. I have not been able to find even a few heatsinks to work with the NMOS350 design in surplus locally. I need to make at least 21 amps for a HT 3-way active crossover based system. Though I have read recently, posted some time ago, that a 4-way active is better for a number of reasons. I will have to research the difference as it is just not cost of amps, but a 4-way crossover active or not is far more complex to execute. There is little difference in building a 3 or 4-way active crossover.
Even though you did not use this thread as a point of reference and mistakenly posted to the thread I still think the design elements you used are of value to this thread. Sometimes a mistake happens to not be such mistake after all. At least in this case and in my opinion. If a moderator decides to pull your posing to another thread I sure hope they replace your posting in this thread with a link. there have already been a number of clearly off thread topic postings to this thread and other threads I have read and they were not pulled to a different thread. If you open a new thread or a moderator does, pulling these postings or not, I for one would like to know if you opened a new thread on your design.
Your insight about "quasi complementary designs might be more subject to asymmetrical clipping" and checking your symmetrical design without the diodes which confirmed the clipping was still symmetrical was very helpful.
Based on your comments for the TLP regulated and the are the same voltage as the output stae enlightens a previous idea I wanted to test. I will now do so at the same rail voltage and of course regulated. This may be the best compromise for the quasi complementary design of the quasi NMOS. I will likely then use a seperate bridge/filter capacitor for the driver stage.
Based on your PSU topology I would like to clarify your deisgn PSU in summary the LTP is a regulated +/-55, the VAS/Driver stage is +/-78V and the output stage is +/-65V? Assuming one has an output rail voltage of +/-X volts, how would you calculate what you would use for the LTP and VAS/Driver stage voltages as Y and Z respectively? I also have to assume one of more of these voltages factor into the amount of gain you design and how you determine the gain you set the amplifier to? I may be incorrectly assuming your add in a "safety margin" of the gain vs the PSU supply voltages? Then again maybe I should not assume 😉
My expereinces with LTspice were not great. LTspice seems to run well under the WINE emulator of Linux. A few problems exist, mostly related to selecting a part. The most important part and for reasons unknown the WINE emulator has some problems with this basic Windows function. The WIME emulator issues aside LTspice had some really basic problems. The first was when I entered the quasi NMOS design in LTspice from scratch and he models of parts not in LTspice. LTspice could not simulate or work with it at all. When I voiced that experience on this thread a gent posted his LTspice of the NMOS350. To that point I was never aware anyone had put quasi's NMOS350 into LTspice. His was the same as mine, except his was missing a few parts that were optional, some of the other values he used were different. Even so and with me adding in the few optional parts and normalizing the parts values LTspice could simulate the NMOS350. At least what I thought until I made a really gross value change to a key resistor that I am told should cause the TLP not to work. Well LTspice still simlated the design with the gross value change in a LTP resistor as if nothing had been changed! Not good as that resistor I wanted to fine tume for the rail rail voltages I wanted to test with. For reasons I will not understand LTspice seems not to understand the importance of the LTP resistor so I stopped using LTspice. Hence I am back to square zero in finding a ECAD/GUI based SPICE program that I can use and works using Linux.
I can understand why you feel and like a fully complementary amplifier design. For me the reasons for choosing a N-Channel design were it is hard for me to obtain complementary BiPolar ouput devices, notwithstanding the fact they are more expensive even for my modest power needs. Too many problems with counterfeit BiPolar output devices. I thouh I bought the genuine set of devices to build some BiPolar amps, but one side of the apir was very suspect. I do not have a RLC bridge to make a definitve conslusion. I seem to like the more robust nature of MOSFETs which also avoids the secondary breakdown voltage character of BiPolars. Thre are not as many complementary MSOFETs devices, let alone availability for me. There are alot of N-Channel MOSFETs I can purhcase locally. Also for right or wrong reasons, BiPolar or MOSFET I like the idea of the same output device on the + and - rail.
I am aware of the laterial 2SK/2SJ and ECX/ECN MOSFET devices. I was not aare of the BUZ901/906 being laterial devices. As much as I might be interested in using laterial MOSFETs, even at modest power handing abilities I need I just cannot afford these devices for the number of amplifiers I need to build. If I win the lotery well that would be a different story! lol lol lol I could afford and I still build my own amps and still build the quasi amps.
That said I was lucky about 8 months ago to come across some commerical receivers that have toroids. use the 2SC5200/2SA1943 output devices and 2SD669A/2SB649A drivers. What is so great was they all had a jumper to allow one to bypass the internal preamp, full speaker binding posts and at a price that beat all the brand names ignoring the china knock off devices they were using. These amps have excellent protection circuits and a inrush power on delay as well. Price was $80.00 CND! I knew there was a 6 channel HT version, but could not find one not the store carrying the stero version carry the HT version for reasons unknown.. Then I found a HT version and picked it up. Aside from the features I noted for the stereo it also had a dedicated 6 channel input which is great match for my Yahaha DSP-3000. this means I have a reference for how a Bipolar amp will sound, and once I build some quasi NMOSs I will be able to compare.
If I do get the time the best comparison would be to build the Brother of Quasi and then compare as this would be basically comparing the same amplifier, one with MOSFET outputs and one with BiPolar. Quasi even has a Class A in the wings of a similar design and uses BiPolar outputs. (Hint, maybe quasi will design a MOSFET output version of the Class A.)
Regards,
John L. Males
Willowdale, Ontario
Canada
01 January 2008 (12:00 - 12:30)
01 January 2008 (14:00 -) 16:17
01 January 2008 16:39 Typo corrections
Official Quasi Thread Researcher
jethari said:
Hi Harri,
The best description regarding the details and meanings of the light bulb behaviour I found were in the Initial Checkout: section. I have copy/pasted the text of that section for reference should that link ever disappear:
Regards,
John L. Males
Willowdale, Ontario
Canada
01 January 2008 (16:39 -) 16:52
Official Quasi Thread Researcher
John
I'll try the light bulb method. But for your info the mosfets are stone cold under no signal conditions. I'm using an ipod as the input and the moment I pause the music or even reduce the input volume, the mosfets start to cool down.
Hari
I'll try the light bulb method. But for your info the mosfets are stone cold under no signal conditions. I'm using an ipod as the input and the moment I pause the music or even reduce the input volume, the mosfets start to cool down.
Hari
jethari said:
Hi Hari,
I know nothing about an iPod other than what I see of the public using them.
Just make sure you do all the tests with the input shorted. If possible nothing connected to the output of the amp. If the MOSFETs are stone cold when there is no input my guess will be that the amp will pass the light bulb tests, unless you observe the Glow-Dim-Glow-Dim result of the test.
Have you tried connecting your amps to a known quality preamp or directly to a known quality CD player to see if the behaviour you have experienced is same or different?
I am assuming you set the bias and offset per quasi's Setup Guide instructions? If so, did you encounter any issues when performing the Setup Guide instructions?
I forgot to note the light bulb is in series with the Hot side of the AC line.
Regards,
John L. Males
Willowdale, Ontario
Canada
01 January 2007 (19:35 -) 19:45
Official Quasi Thread Researcher
Vivek said:
Hi Vivek,
First check what DC voltage rails you have with the light bulb in series and let us know.
The other thing you could do is check whether the amp is actually amplifying by connecting an input signal and checking if you get an output.
Once we have the DC voltage rails we'll go from there.
jethari said:Seasons greetings to all of you
I finally got my NMOS 350 going. Came up with some problems. Even with a low bias current (15ma) the fets are getting very hot at medium to high music levels. I've tried it with both two pairs as well as one pair. My supply rails are +/- 65V dc. In fact at normal bias current levels the fets just blew at high music levels (I have now a collection of blown fets
Right now I'm running one pair at 15ma bias. Though the sound is simply brilliant (the bass goes real low and clean), the heating of the fets worries me. I am not driving it any higher.
If you recall, I'm using the NMOS 200 board layout for the NMOS 350 circuit. Could this have any bearing on my problem?
Regarding T8, is it supposed be thermally tracking the output pairs? If so would it be prudent to use a MJE340 instead? Also, in the NMOS 200 schematic there is a 0.1uf cap across the emitter-collector of T8 but not in the NMOS 350 schematic. I've included this in my layout. Could this have any bearing?
Hari
Gday Hari,
Can you please post a photo showing the mounting of the output FETs. It could be that you are not removing the heat well enough. On the original Nmos350 the FETs are mounted directly onto a large heatsink.
If you have access to a CRO and a signal generator have a look at the output for possible oscillations.
Cheers fellas
Q
jethari said:
Hari,
Sorry I forgot to answer this question you asked in my prior replies. The 0.1uF cap across emitter-collector of T8 is in fact supposed to on the NMOS350/500. The capacitor has always been on the NMOS350 PCBs since the PCB was first issued, but was missed on all but the last NMOS350 schematic posted. The NMOS350 schematic quasi currently has posted on his website is missing the 0.1uF across T8's E-C as it was the only schematic quasi placed the currents on prior to the last correction to the schematic which was to include the mising 0.1uF across E-C of T8 on the schematic.
Regards,
John L. Males
Willowdale, Ontario
Canada
01/02 January 2008 (23:25 -) 00:15
Official Quasi Thread Researcher