Reply on my MATCHFET MACHINE
Well, I already knew Nelson's reply.
.....to all answers I have seen,you guys seem to complicate your mind a little.(...)
I thought about this machine just to get a less complicated matching of the VGS.I don't think I need a computer to achieve this.
I don't find matching of mosfets so interesting.If I could have them matched from someone,I would prefer it.But....I prefer having fun listening to the amp.In fact,I sometimes organize critical listening tests just for myself,early in the morning to have fresh ears 4'o'clock is fine,try it and you'll find yourself a more critical listener;at this time ambiant noise is very low,allowing a more precise opinion.
.......but most of the time I prefer "pleasure" listening.
To go back to the mosfets match discussion,I don't think it makes so difference for only sound to match mosfets with such precisemania.The Vgs will change with temperature of course,but...if matched under same conditions,I believe the respective VGS will change same way once onto an amp's heatsink...
Well, I already knew Nelson's reply.
.....to all answers I have seen,you guys seem to complicate your mind a little.(...)
I thought about this machine just to get a less complicated matching of the VGS.I don't think I need a computer to achieve this.
I don't find matching of mosfets so interesting.If I could have them matched from someone,I would prefer it.But....I prefer having fun listening to the amp.In fact,I sometimes organize critical listening tests just for myself,early in the morning to have fresh ears 4'o'clock is fine,try it and you'll find yourself a more critical listener;at this time ambiant noise is very low,allowing a more precise opinion.
.......but most of the time I prefer "pleasure" listening.
To go back to the mosfets match discussion,I don't think it makes so difference for only sound to match mosfets with such precisemania.The Vgs will change with temperature of course,but...if matched under same conditions,I believe the respective VGS will change same way once onto an amp's heatsink...
As temperature changes, Vgs changes relative to each other. If this wasnt the case, precise matching would be a whole lot easier.... sure we probably dont need to match them this closely but hey if you've got the time and enough devices on hand why not? The smaller we can get each possible variable in the amp, the more consistent the sound quality will be between amps. It's all a matter of how far your willing to take it.
Board Trace and Cable Resistance
Since it's so critical to have match transistors & there is no adjustment on the board, how critical is the resistance of the output board trace and wire resistance coming from the main board? We're talking about mV here and there should not have any major voltage drops but still, what should be the maximum length of the cable be or the width & thickness of the output board trace be before it mess up the Vgs? Or their effect is unnoticeable ...... because of the way the circuit was designed?
Since it's so critical to have match transistors & there is no adjustment on the board, how critical is the resistance of the output board trace and wire resistance coming from the main board? We're talking about mV here and there should not have any major voltage drops but still, what should be the maximum length of the cable be or the width & thickness of the output board trace be before it mess up the Vgs? Or their effect is unnoticeable ...... because of the way the circuit was designed?
So, if I may summarize what I've read so far:
the Vgs change compared to time isn't linear, and longer times result in less chance of an error and allows the device to stabilize more. There is no great need to wait such a long time, but the time each device is measured for must be closely monitored, especially at shorter times. Correct?
So, if you want the best accuracy and a quick measurement time, why not use a timer that switches on when the voltage is applied to the device? Have it beep or something when it is timed out, and record the measurement at that time. I know I don't want to wait for 5 minutes each device when I'm building my A75's- that's 4 hours worth of matching! (call me impatient). Excuse me if I'm stating the obvious here.
steve
the Vgs change compared to time isn't linear, and longer times result in less chance of an error and allows the device to stabilize more. There is no great need to wait such a long time, but the time each device is measured for must be closely monitored, especially at shorter times. Correct?
So, if you want the best accuracy and a quick measurement time, why not use a timer that switches on when the voltage is applied to the device? Have it beep or something when it is timed out, and record the measurement at that time. I know I don't want to wait for 5 minutes each device when I'm building my A75's- that's 4 hours worth of matching! (call me impatient). Excuse me if I'm stating the obvious here.
steve
the thing is, after only a matter of seconds on the test bench, the Vgs may still be changing by upto about 0.1V per second which will make it extremely difficult to obtain accurate results. Also, not all devices will differ equally over a given period of time so devices that match after a few seconds may not after a few minutes because die temp will have had a chance to stabalise whereas after a few seconds, it is still rising rapidly.
So, the variable is the rate of rise in die temperature, and you MUST let it stabilize? Has anyone looked into how long (or short) of a time makes a significant difference in matching? If you match them, say, in exactly 20 seconds, and then match them in, say, 2 minutes, will the tolerance be <I>significantly</I> different? I guess my question would be what is the minimum time needed to get them matched accurately enough? We (audio DIYers) tend to go overboard on just about everything. I guess this discussion is relatively pointless, as we tend to go to extremes anyway. Why not wait the extra time? I guess to me it's more about learning than anything. I will look at the data sheets for my IRFP240's. I remember seeing different curves for different temps. Maybe there is something that applies here.
Steve
Steve
Beating up a dead horse!
Guys,
Just follow Mr.Pass's advice and you can't go wrong, after all he wrote the book. You are making this way too complicated, I have matched hundreds of devices with no problem.
I have tried it both ways and I rather spend the saved time building something.
On the other hand if spending hours matching devices is your thing, who am I to judge...........
Jam
P.S. Maybe Harry has some sage advice.
Guys,
Just follow Mr.Pass's advice and you can't go wrong, after all he wrote the book. You are making this way too complicated, I have matched hundreds of devices with no problem.
I have tried it both ways and I rather spend the saved time building something.
On the other hand if spending hours matching devices is your thing, who am I to judge...........
Jam
P.S. Maybe Harry has some sage advice.
Mosfet matching
I only match mine on full moon since the gravitational forces of such a large effect. The Mosfets have fairly large source resistances which act as degeneration and tends to equalize the bias currents anyway. I think Mr. Pass probabaly tried to match the output fets so that the current souce DC current matched the output stage to minimize the DC offset variations. There are things to worry about in large scale production that are not a big deal in a one off build. To put things in perspective: for a 0.5 ohm source resistor at 1 amp, the tolerence for a 5% part will give variations of 50mV in the effective Vgs matching. Why not put your efforts into close matching of the input differentional pair which will have much greater impact on distortion and DC offset. The Fet matching disscussions are becoming as stale as the heat sink hand wringing.
I only match mine on full moon since the gravitational forces of such a large effect. The Mosfets have fairly large source resistances which act as degeneration and tends to equalize the bias currents anyway. I think Mr. Pass probabaly tried to match the output fets so that the current souce DC current matched the output stage to minimize the DC offset variations. There are things to worry about in large scale production that are not a big deal in a one off build. To put things in perspective: for a 0.5 ohm source resistor at 1 amp, the tolerence for a 5% part will give variations of 50mV in the effective Vgs matching. Why not put your efforts into close matching of the input differentional pair which will have much greater impact on distortion and DC offset. The Fet matching disscussions are becoming as stale as the heat sink hand wringing.
MATCHMANIA part 2
I totaly agree with jam. It is better time spent on other parts of the project,such as drilling,tappering,make the out look finished,ensure wires are short and tightened/soldered well.
I think cables link to power mosfets should be same lengh and as short as possible. As I always use full copper/one strand 2.5mm2 wire,I think resistance problems to the high current are not so important. For inductance,same way. The capacitance problem is there,but the design of the Alephs makes it no important,as the bandwith of the amp. finds a natural cut/off with the input mosfet capacitance,as we don't go for a 1 Mhz freq.response,I think we can safely throw away our fears or doubts.
Even Nelson stated that he wouldn't give his own circuit board,as it would add nothing new on the table.
But perhaps someone made tests with an Aleph to evaluate scientifically what occurs in extreme disfavorable situations?
I totaly agree with jam. It is better time spent on other parts of the project,such as drilling,tappering,make the out look finished,ensure wires are short and tightened/soldered well.
I think cables link to power mosfets should be same lengh and as short as possible. As I always use full copper/one strand 2.5mm2 wire,I think resistance problems to the high current are not so important. For inductance,same way. The capacitance problem is there,but the design of the Alephs makes it no important,as the bandwith of the amp. finds a natural cut/off with the input mosfet capacitance,as we don't go for a 1 Mhz freq.response,I think we can safely throw away our fears or doubts.
Even Nelson stated that he wouldn't give his own circuit board,as it would add nothing new on the table.
But perhaps someone made tests with an Aleph to evaluate scientifically what occurs in extreme disfavorable situations?
Source Resistors
Jam, it was the source resistors that I was talking about! The point is that the paralleled output Fets have a voltage across them that is the sum of Vgs plus the voltage across the source resistor due to the bias current. Differences in the source resistors' value will give a slightly different bias current for a given voltage from the gates to the negative supply. For two Mosfets with source resistors, in parallel, the bias current can be slightly different due to different values of the source resistors, even if thier Vgs is identical when matched at the same bias current. The source resistances act to equalize the current between devices which is the usually the reason they are included in an output circuit. Paralleled Mosfets without source resistors would require very close matching to insure that they shared the current equally. The fairly large values of the source resistors actually make device matching less critical than what one would think at first glance. I actually match devices with the source resistor in place for each device. This also serves to minimize themal drift at operatinf temperature.
Jam, it was the source resistors that I was talking about! The point is that the paralleled output Fets have a voltage across them that is the sum of Vgs plus the voltage across the source resistor due to the bias current. Differences in the source resistors' value will give a slightly different bias current for a given voltage from the gates to the negative supply. For two Mosfets with source resistors, in parallel, the bias current can be slightly different due to different values of the source resistors, even if thier Vgs is identical when matched at the same bias current. The source resistances act to equalize the current between devices which is the usually the reason they are included in an output circuit. Paralleled Mosfets without source resistors would require very close matching to insure that they shared the current equally. The fairly large values of the source resistors actually make device matching less critical than what one would think at first glance. I actually match devices with the source resistor in place for each device. This also serves to minimize themal drift at operatinf temperature.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.