Status report and matching of FET's result
I have gotten PCB's and expect to start building input stages this weekend. Sorry about taking so long, but I must say my experience with home PCB making and etching was not a good one. Pro's do it better -- thanks!
Today I matched my batches of IRF610 and IRF9610. Each population was 30 units. I am not sure whether they were all from the same lot codes (I did not bother to check this)
Out of each lot, I was able to select 10 pairs that were all matched to within 5mV at the chosen 30mA (clue -- look at the diagram on page 1 ....), no problem!
Out of the p-types I got 7 pairs to less than 1mV. Range was .18V, SD: 0.057V, Mean 3,52V
Out of the n-types I also got 7 pairs to less than 1mV. Range was .37V, SD: 0.089V, Mean 3,92
I realize this is all somewhat academic since I was not able to control temperatures perfectly, but it demonstrates the point that pairs are easy to match up. Besides, these units were inexpensive.
I am eager to test my output transistors which will be more interesting since I am looking for larger sets, but I also have a population of 60.
I have gotten PCB's and expect to start building input stages this weekend. Sorry about taking so long, but I must say my experience with home PCB making and etching was not a good one. Pro's do it better -- thanks!
Today I matched my batches of IRF610 and IRF9610. Each population was 30 units. I am not sure whether they were all from the same lot codes (I did not bother to check this)
Out of each lot, I was able to select 10 pairs that were all matched to within 5mV at the chosen 30mA (clue -- look at the diagram on page 1 ....), no problem!
Out of the p-types I got 7 pairs to less than 1mV. Range was .18V, SD: 0.057V, Mean 3,52V
Out of the n-types I also got 7 pairs to less than 1mV. Range was .37V, SD: 0.089V, Mean 3,92
I realize this is all somewhat academic since I was not able to control temperatures perfectly, but it demonstrates the point that pairs are easy to match up. Besides, these units were inexpensive.
I am eager to test my output transistors which will be more interesting since I am looking for larger sets, but I also have a population of 60.
McKaj,
you should be very careful about posting this type of things on a public forum. Much like underage drinking, smoking weed, and your parents having sex this is something that everyone knows it happens and everyone does it (but me, the illegal stuff that is) but you just don't talk about it. And if you don't know about it means you shouldn't be told.
Itsy bitsy piece of advice: remove the post.
you should be very careful about posting this type of things on a public forum. Much like underage drinking, smoking weed, and your parents having sex this is something that everyone knows it happens and everyone does it (but me, the illegal stuff that is) but you just don't talk about it. And if you don't know about it means you shouldn't be told.
Itsy bitsy piece of advice: remove the post.
Patience ...
My apologies for the lack of progress on this thread.
Current status is that I have but for some missing parts (resistors) that I will get today or tomorrow, I have completed input stages for two channels. I have not done any PSU work yet, but this is going to be more traditional anyway and should not prove to be a problem. Since I made my PCB so tight, it has been somewhat of a challenge to fit all the parts. I had to revert to placing components on both sides of the PCB -- some of this was planned for, but not all, and without plated through holes ... I also had not anticipated that form factor of high quality components such as capacitors is not always a choice (OS con's for example, not available in radials). I am still quite happy with the layout.
My hope is to make available simple measurements (gain, frequency response) before July 4, but I am not at all certain that I will make this deadline. If I do miss it, there will be a delay until July 16 or so, but then I might be able to make distortion measurements as well.
Bear in mind this is for input stage with dummy load only.
If the results are good, I should be able to have the amps fully built relatively quickly.
Petter
My apologies for the lack of progress on this thread.
Current status is that I have but for some missing parts (resistors) that I will get today or tomorrow, I have completed input stages for two channels. I have not done any PSU work yet, but this is going to be more traditional anyway and should not prove to be a problem. Since I made my PCB so tight, it has been somewhat of a challenge to fit all the parts. I had to revert to placing components on both sides of the PCB -- some of this was planned for, but not all, and without plated through holes ... I also had not anticipated that form factor of high quality components such as capacitors is not always a choice (OS con's for example, not available in radials). I am still quite happy with the layout.
My hope is to make available simple measurements (gain, frequency response) before July 4, but I am not at all certain that I will make this deadline. If I do miss it, there will be a delay until July 16 or so, but then I might be able to make distortion measurements as well.
Bear in mind this is for input stage with dummy load only.
If the results are good, I should be able to have the amps fully built relatively quickly.
Petter
OK then!
Status: Working prototype of input stage!!!!
(This means that 90% of the work is done -- this is the innner loop of the drawing in the first post of this thread, and the only section requiring adjustment and real design work, a breakthrough that has been coming for a lOOOONG time now).
After 2 days in the lab, it turns out that my prototyping board is working after all! The identical sister is not working, probably a blown FET or fried resistors -- I did have some smoke in a resistor set. When you are working with +-50V supplies (and no ground), it is quite easy to fry things. Very strange that I did not fry more stuff. I did choose to start of with my best matched FET's and as such I was relatively brave (or foolish). Due to space limitations (overoptimization) and very very high voltage rails (by choice) I really would want a little more cooling. If I fail that, I will add a small fan (to the prototype set).
So, I have sinusoidal balanced in, with sinusoidal balanced out, and significant gain. About 60V pk-pk output with "configurable behaviour" -- add more voltage, get more output swing. This is a one-gainstage design. Output stage will just provide power. When you read this, you now know what I have been aiming for -- a configurable power amp. Add more juice to the output stage and suck out more power, for example 4 times as much. 99% of the time you are in low power mode with low heat + high class A drive in output stage. 1% of the time you are in high power mode with low class A drive ("low" compared to commercial products with normal price tags).
I have not tested with load yet, but hooking up to a power stage will be a lot of fun. I noticed that i had dropoff in gain at around 10KHz, much too low. I really should consult with the specs of the originals for comparison.
I also need to verify stability -- the circuit needs to be completely in balance. A word of construction advice, use mathed devices in equivalent roles -- ie bottom 4 current sources match up, top two match up -- inverted cascode probably not important but why not -- input devices a resounding YES! your best pair.
When building, set up current in top sources first with a high quality current meter measured after the inverted cascode. Then set up bottom sources to half that value -- before stuffing board completely. Complete and adjust outer lower pair for correct output voltage -- will probably be way out of balance and relatively insensitive and suddenly swing over really fast -- this is actually in line with my simulations that you may recall I had trouble with once I switched to real devices.
Also when building, use a triplet of resistors to set up all your trimpot turns (15 in my case) to be in your desired window (1 resistor drawing most of the current in paralells with a trimpot + a series resistor). I have about 1 turn per .08mA on the current sources ... I have an Excel spreadsheet which uses linear optimization methods to solve this (untrivial) problem. I call this a less than trivial problem because I gave up after 40 minutes with pen and paper and did the job in 5 minutes in Excel .... Of course there are 2 solution sets and 3 unknowns -- when I was in college, I probably would have done it for the exercise. Now I have other things to do.
Now, I need a place to host the pictures so I can put them up on this board. Anyone care to host it for me and provide a URL? Nothing incriminating here ... Send me an email through the system.
Petter
Status: Working prototype of input stage!!!!
(This means that 90% of the work is done -- this is the innner loop of the drawing in the first post of this thread, and the only section requiring adjustment and real design work, a breakthrough that has been coming for a lOOOONG time now).
After 2 days in the lab, it turns out that my prototyping board is working after all! The identical sister is not working, probably a blown FET or fried resistors -- I did have some smoke in a resistor set. When you are working with +-50V supplies (and no ground), it is quite easy to fry things. Very strange that I did not fry more stuff. I did choose to start of with my best matched FET's and as such I was relatively brave (or foolish). Due to space limitations (overoptimization) and very very high voltage rails (by choice) I really would want a little more cooling. If I fail that, I will add a small fan (to the prototype set).
So, I have sinusoidal balanced in, with sinusoidal balanced out, and significant gain. About 60V pk-pk output with "configurable behaviour" -- add more voltage, get more output swing. This is a one-gainstage design. Output stage will just provide power. When you read this, you now know what I have been aiming for -- a configurable power amp. Add more juice to the output stage and suck out more power, for example 4 times as much. 99% of the time you are in low power mode with low heat + high class A drive in output stage. 1% of the time you are in high power mode with low class A drive ("low" compared to commercial products with normal price tags).
I have not tested with load yet, but hooking up to a power stage will be a lot of fun. I noticed that i had dropoff in gain at around 10KHz, much too low. I really should consult with the specs of the originals for comparison.
I also need to verify stability -- the circuit needs to be completely in balance. A word of construction advice, use mathed devices in equivalent roles -- ie bottom 4 current sources match up, top two match up -- inverted cascode probably not important but why not -- input devices a resounding YES! your best pair.
When building, set up current in top sources first with a high quality current meter measured after the inverted cascode. Then set up bottom sources to half that value -- before stuffing board completely. Complete and adjust outer lower pair for correct output voltage -- will probably be way out of balance and relatively insensitive and suddenly swing over really fast -- this is actually in line with my simulations that you may recall I had trouble with once I switched to real devices.
Also when building, use a triplet of resistors to set up all your trimpot turns (15 in my case) to be in your desired window (1 resistor drawing most of the current in paralells with a trimpot + a series resistor). I have about 1 turn per .08mA on the current sources ... I have an Excel spreadsheet which uses linear optimization methods to solve this (untrivial) problem. I call this a less than trivial problem because I gave up after 40 minutes with pen and paper and did the job in 5 minutes in Excel .... Of course there are 2 solution sets and 3 unknowns -- when I was in college, I probably would have done it for the exercise. Now I have other things to do.
Now, I need a place to host the pictures so I can put them up on this board. Anyone care to host it for me and provide a URL? Nothing incriminating here ... Send me an email through the system.
Petter
Hi You can host the pics at my web site if you wish.
regards
Anthony Holton
http://www.aussieamplifiers.com
regards
Anthony Holton
http://www.aussieamplifiers.com
How similar to the original
It is hard to tell, but judging from the information I have up to this point as well as comments from Nelson Pass in this very thread, I suspect it is almost completely the same.
Read what NP said and judge for yourself -- last post on page 3.
Photos will be forthcoming when I can get hold of a digital camera.
Petter
[Edited by Petter on 07-26-2001 at 05:20 PM]
It is hard to tell, but judging from the information I have up to this point as well as comments from Nelson Pass in this very thread, I suspect it is almost completely the same.
Read what NP said and judge for yourself -- last post on page 3.
Photos will be forthcoming when I can get hold of a digital camera.
Petter
[Edited by Petter on 07-26-2001 at 05:20 PM]
Except for the small fact that Nelson uses 80 output devices instead of 16......The X-1000 dissipates about 600W per channel, so if you would have used only 16 output devices they would have a real tough job. Spreading the load, as Pass does, is the only way for these kind of high-powered and hot-running amps to last....(please bare this in mind)
Gordon Gekko
Gordon Gekko
I have uploaded Petter's pictures to this url (thank you Petter!)
http://www.diyaudio.com/projects/petter
[Edited by Jason on 07-30-2001 at 12:57 AM]
http://www.diyaudio.com/projects/petter
[Edited by Jason on 07-30-2001 at 12:57 AM]
What are these files
Triplets: An Excel optimization model for creating compound resistors where current in trimmer is minimal, all turns of trimmer are used in desired window, and progressions is relatively linear.
Other files: Images of circuit. Don't laugh at the amateurish look + crammed board ...
Please note:
1. Brass screws (no ferric materials here!)
2. Small dimensions of board
3. Lack og ground
4. Power supply comb pattern is "filler" to fill the board. I figured if I did this I would be able to find use for it.
5. Rounded features in PCB track corners.
Also note what I did wrong:
1. PCB tracks could be thicker
2. Size could be made a little bigger
3. Mounting holes should be set up (not really a problem, but ...)
4. I ended up using larger resistors than expected (which is good because they don't modulate so easily but the pattern was a little tight for them.
5. Trimmers are too close to other things.
6. Pads could be a little bigger.
7. Transistors have a thicker side. If you put two of them together, that is about 8 mm of space which I did not plan for! Check out the central right transistor (top) and you will see what I mean!
8. Small silly things that amateurs do.
But hey, this is my first print. There are also other modifications that could be done.
Petter
Triplets: An Excel optimization model for creating compound resistors where current in trimmer is minimal, all turns of trimmer are used in desired window, and progressions is relatively linear.
Other files: Images of circuit. Don't laugh at the amateurish look + crammed board ...
Please note:
1. Brass screws (no ferric materials here!)
2. Small dimensions of board
3. Lack og ground
4. Power supply comb pattern is "filler" to fill the board. I figured if I did this I would be able to find use for it.
5. Rounded features in PCB track corners.
Also note what I did wrong:
1. PCB tracks could be thicker
2. Size could be made a little bigger
3. Mounting holes should be set up (not really a problem, but ...)
4. I ended up using larger resistors than expected (which is good because they don't modulate so easily but the pattern was a little tight for them.
5. Trimmers are too close to other things.
6. Pads could be a little bigger.
7. Transistors have a thicker side. If you put two of them together, that is about 8 mm of space which I did not plan for! Check out the central right transistor (top) and you will see what I mean!
8. Small silly things that amateurs do.
But hey, this is my first print. There are also other modifications that could be done.
Petter