Scott,
Yikes, a modification to my sacred circuit! Whatever shall I do? Let me run in circles for a minute or two until my karma gets rebalanced.
Okay, I'm over it.
Actually, I think I remember seeing the version you're referring to, but I don't recall how they had it set up. Therefore, in my best time-honored manner, I'll wing it.
If, for whatever reason, Zeners bother you, here are two more ideas:
--Series two (or more) voltage references. If you can get 9V, great. Otherwise, stack two 5V for instance (i.e. 10V), and use a voltage divider or pot to set the voltage exactly where you want it.
--Don't worry about finding a 9.1V reference...use your 5mA current source across an 1820 ohm resistor. Presto! Instant 9.1 volts.
And a third idea, just for fun:
--If you've got, say, a 6V reference, then simply recalculate the resistors under the tail end of the current source. You'll need something more than the Vgs of the MOSFET, but once you get over that limit, it's a simple E/R=I calculation to set the current.
Think creative thoughts, man. You'll get there.
Grey
Yikes, a modification to my sacred circuit! Whatever shall I do? Let me run in circles for a minute or two until my karma gets rebalanced.
Okay, I'm over it.
Actually, I think I remember seeing the version you're referring to, but I don't recall how they had it set up. Therefore, in my best time-honored manner, I'll wing it.
If, for whatever reason, Zeners bother you, here are two more ideas:
--Series two (or more) voltage references. If you can get 9V, great. Otherwise, stack two 5V for instance (i.e. 10V), and use a voltage divider or pot to set the voltage exactly where you want it.
--Don't worry about finding a 9.1V reference...use your 5mA current source across an 1820 ohm resistor. Presto! Instant 9.1 volts.
And a third idea, just for fun:
--If you've got, say, a 6V reference, then simply recalculate the resistors under the tail end of the current source. You'll need something more than the Vgs of the MOSFET, but once you get over that limit, it's a simple E/R=I calculation to set the current.
Think creative thoughts, man. You'll get there.
Grey
Grey's current source design?
Scott,
An LM 329 running at 1 ma is fine. The larger current the jfet CCS diode the lower its impedance. You will get much more power supply rejection for the LM 329 voltage reference with the 1 mA current jfet. Don't forget a 1K gate resistor for the ZVP3310. Setting 10 volts across the device is a 150 mW which is no problem for dissipation and helps to lower the drain to source capacitance for the current source. Don't hesitate to put a series resistor between the drain of the ZVP3310 and the sources of the mosfet differential pair to limit the voltage across the drain to source of the current source. It will help to isolate capacitance a little more and detracts nothing from the performance of the current source. I really believe this current source approach is worth implementing since it has several orders of magnitude better rejection and noise. A resistor biased zener with low voltage rails has pretty poor supply rejection and is pretty noisy. Sorry if I clipped anyone's wings............
Scott,
An LM 329 running at 1 ma is fine. The larger current the jfet CCS diode the lower its impedance. You will get much more power supply rejection for the LM 329 voltage reference with the 1 mA current jfet. Don't forget a 1K gate resistor for the ZVP3310. Setting 10 volts across the device is a 150 mW which is no problem for dissipation and helps to lower the drain to source capacitance for the current source. Don't hesitate to put a series resistor between the drain of the ZVP3310 and the sources of the mosfet differential pair to limit the voltage across the drain to source of the current source. It will help to isolate capacitance a little more and detracts nothing from the performance of the current source. I really believe this current source approach is worth implementing since it has several orders of magnitude better rejection and noise. A resistor biased zener with low voltage rails has pretty poor supply rejection and is pretty noisy. Sorry if I clipped anyone's wings............
GRollins said:
I'm getting confused now. Might be because of the shear size of the original thread, but most likely because I don't know much about power supplies in general. Sorry, but I just can't seem to put all this info together.
I was using the AXE-1 spreadsheet. If I have a dual secondary 16V transformer and use both secondaries for one channel, do I plug enough rail voltage in the "B12" cell to arrive at ~16V or ~30V in the "B23" cell?
What about if I use one 16V secondary per channel? Probably not enough voltage to make the Mosfets work.
Netlist said:
If you mean CEL B32 that would be R5/R6 and R40/41.
The calculation is based on the 0.5V that should be over these resistors. Keep in mind this is only a theoretical value.
Quote from Grey: "There's nothing magic about .5V--you're free to set it where you like."
/Hugo
So Cell B33 is the Resistor value I choose to put in for R5/R6 & R40/R41? So B32 is the theoretical and B33 is the practical?
So if I jumper these resistors, can't I take the output for my FET boards directly from the FET Pads? Then I would only need one other wire for the current sense line to each board.
Anthony
Good Enough?
Here are some measurements from my Aleph X.
14V rails .5V across R5
Q1 3.58 Vgs - Q10 3.51 Vgs
Q2 3.80 Vgs - Q11 3.84 Vgs
Diff Pair 3.693 and 3.679 Vgs Q6a 3.87 Vgs
Abs. Offset <50 mV
DC Offset 180mV - 200mV
I guess i'll have to keep popping in FETs to the Q6a slot until i get
a <100mV DC Offset reading. It's just hard to keep desoldering
that piece. I didn't realize that the diff pair had to match to Q6.
or can one live with the DC offset i'm getting.
thanks!
m.
Here are some measurements from my Aleph X.
14V rails .5V across R5
Q1 3.58 Vgs - Q10 3.51 Vgs
Q2 3.80 Vgs - Q11 3.84 Vgs
Diff Pair 3.693 and 3.679 Vgs Q6a 3.87 Vgs
Abs. Offset <50 mV
DC Offset 180mV - 200mV
I guess i'll have to keep popping in FETs to the Q6a slot until i get
a <100mV DC Offset reading. It's just hard to keep desoldering
that piece. I didn't realize that the diff pair had to match to Q6.
or can one live with the DC offset i'm getting.
thanks!
m.
The diff pair does NOT have to match Q6. Q6 lives in it's own little world, and doesn't care at all what any of the other transistors are up to. It is the Q5-Q7 pair which must be matched to give you a low DC offset between the speaker terminals.
This mismatch is the source of your DC offset: "Diff Pair 3.693 and 3.679 Vgs"
Hope that helps, and saves you some trouble.
This mismatch is the source of your DC offset: "Diff Pair 3.693 and 3.679 Vgs"
Hope that helps, and saves you some trouble.
CCS Continued.
Trying to be creative, without getting too far off of the purpose of this thread.
I greatly appreciate the input from both Grey and Fred. Your inputs are greatly helping me understand how the circuit works.
For grins and giggles, lets take this one step further, lets double the bias of the diff pair. ( More Bias is better right?) If I understand this correctly R23, R25 would go from 392ohm to 195ohm so the absolute DC offset track closer to zero and the gain remains the same. The resistor network (R24,R26,VR2) would have to be adjusted to provide control of Q6 at just above its threshold. Depending on the device used the, network would need to control somewhere around 40 or 70 ohms using a 6.9V reference voltage.
If the irf9610 was used for q6, power dissipation wouldn't be an issue, however with the ZVP3310A option, R48 may have to be adjusted slightly to keep power dissipation in a safer operating range possibly with the sacrafice of some linearity. Based on Freds Comments if I was willing to live with 300mW on q6 R48 could actually stay the same.
The only other issue that I can see from my very simplistic point of view, is that if R48 was used at a sufficient value to keep the Vds of q6 around 10V then a 1/2 watt resistor would be needed. Additionally the power dissipated by the diff pair Q5/Q7 would be somewhere above 0.75 watt meaning that a small sink wouldn't be a bad idea providing they are thermally coupled.
Having said that, are there other issues that pop up when upping the bias of the differential pair?
Thanks
Scott
Trying to be creative, without getting too far off of the purpose of this thread.
I greatly appreciate the input from both Grey and Fred. Your inputs are greatly helping me understand how the circuit works.
For grins and giggles, lets take this one step further, lets double the bias of the diff pair. ( More Bias is better right?) If I understand this correctly R23, R25 would go from 392ohm to 195ohm so the absolute DC offset track closer to zero and the gain remains the same. The resistor network (R24,R26,VR2) would have to be adjusted to provide control of Q6 at just above its threshold. Depending on the device used the, network would need to control somewhere around 40 or 70 ohms using a 6.9V reference voltage.
If the irf9610 was used for q6, power dissipation wouldn't be an issue, however with the ZVP3310A option, R48 may have to be adjusted slightly to keep power dissipation in a safer operating range possibly with the sacrafice of some linearity. Based on Freds Comments if I was willing to live with 300mW on q6 R48 could actually stay the same.
The only other issue that I can see from my very simplistic point of view, is that if R48 was used at a sufficient value to keep the Vds of q6 around 10V then a 1/2 watt resistor would be needed. Additionally the power dissipated by the diff pair Q5/Q7 would be somewhere above 0.75 watt meaning that a small sink wouldn't be a bad idea providing they are thermally coupled.
Having said that, are there other issues that pop up when upping the bias of the differential pair?
Thanks
Scott
Its funny that Nelson says .01 volt is fine...he's got to be talking output devices...not input. I have a pair for Q-5,7 that are within .007 V of each other. I should have this thing up and running tommrrow night so it'll be interesting to see what my offset actually is. I feel that actual subbing of just one of these devices is probably the best way find a match that achieves really low DC offset.
Mark
Mark
Chad, thanks for the info! I have been reading and re-reading
the WIKI too... 1st time builders should really be taking adavantage
of all the hard work that went in to setting that up... great info!
i set up a jig to be able to swap out one of the differential pairs
in circuit. Finding a good match this way takes longer, but for
some reason i feel a little better that it's being done in it's operating
environment. Of course this means they're not mounted
back to back during testing...
wish me luck!
the WIKI too... 1st time builders should really be taking adavantage
of all the hard work that went in to setting that up... great info!
i set up a jig to be able to swap out one of the differential pairs
in circuit. Finding a good match this way takes longer, but for
some reason i feel a little better that it's being done in it's operating
environment. Of course this means they're not mounted
back to back during testing...
wish me luck!
I was also looking to up the bias on the diff pair.What's the best way to do this?How much are they running at now?I thought lowering the 392r would lower the gain?
Another thing about Aleph's and A-X that I don't quite understand - if you increase the ac current gain for better performance into 4ohms what does that mean that for class a operation?Does it become a kind of heavily biased class AB? When does it leave class a operation - when the output stage bias limit is reached or sooner?
Another thing about Aleph's and A-X that I don't quite understand - if you increase the ac current gain for better performance into 4ohms what does that mean that for class a operation?Does it become a kind of heavily biased class AB? When does it leave class a operation - when the output stage bias limit is reached or sooner?
Pass DIY Addict
Joined 2000
Paid Member
Hugo: Thanks for the reply. I went back and re-read the entire thread on the high-powered version. Great stuff!
Grey: I have 2 heatsinks from the original group purchase, and am interested in building a relatively low powered version with dual mono in the same chassis. I have been playing with the AXE-1 spreadsheet and am looking at 15v rails, 3A bias with 4 fets per channel. Seems to require about 90w dissipation per channel making it a good fit to the heatsinks.
I have also been considering the logistics of putting 3 channels in the same chassis (front sound stage for my home theater system), but don't think it will all fit in an 11" cube...
Grey: I have 2 heatsinks from the original group purchase, and am interested in building a relatively low powered version with dual mono in the same chassis. I have been playing with the AXE-1 spreadsheet and am looking at 15v rails, 3A bias with 4 fets per channel. Seems to require about 90w dissipation per channel making it a good fit to the heatsinks.
I have also been considering the logistics of putting 3 channels in the same chassis (front sound stage for my home theater system), but don't think it will all fit in an 11" cube...
Regarding the front end differential offset vs. the actual offset at the output: You can use the inevitable slight mismatch of the output devices to counteract the front end differences. Just swap them back and forth. With luck, you can find a pair that will yin to the other's yang and produce a nice, harmonious balance.
protos,
Yes, lowering the value of the front end load resistors will lower the gain...if done in isolation. There are two factors that you want to balance--the gain and the DC offset across those resistors. The gain isn't critical, actually. The DC offset is, however. That DC offset sets the bias for the output MOSFETs; switches 'em on.
If you increase the bias so as to handle a lower Z load, you're ensuring that it reamins in class A. The Aleph output stage does not do class AB gracefully. The current source and the output won't drive symmetrically, and the signal gets kinda lopsided.
Eric,
I haven't looked at anyone's spreadsheets, so I can't vouch for accuracy. I've got one I hamered together a couple of years ago when I was torturing the Aleph circuit to see what it would do. I then went back and modified it to handle the Aleph-X. It's ugly as home made sin, but it does what I need, so I've never bothered to look at the others.
Is the 3A bias per side or for the whole amp? How much power were you shooting for and into what load?
An 11" cube sounds pretty limited for heat dissipation for three channels. Perhaps forced air cooling would be an option.
protos,
Yes, lowering the value of the front end load resistors will lower the gain...if done in isolation. There are two factors that you want to balance--the gain and the DC offset across those resistors. The gain isn't critical, actually. The DC offset is, however. That DC offset sets the bias for the output MOSFETs; switches 'em on.
If you increase the bias so as to handle a lower Z load, you're ensuring that it reamins in class A. The Aleph output stage does not do class AB gracefully. The current source and the output won't drive symmetrically, and the signal gets kinda lopsided.
Eric,
I haven't looked at anyone's spreadsheets, so I can't vouch for accuracy. I've got one I hamered together a couple of years ago when I was torturing the Aleph circuit to see what it would do. I then went back and modified it to handle the Aleph-X. It's ugly as home made sin, but it does what I need, so I've never bothered to look at the others.
Is the 3A bias per side or for the whole amp? How much power were you shooting for and into what load?
An 11" cube sounds pretty limited for heat dissipation for three channels. Perhaps forced air cooling would be an option.
Heatsink Source Needed
OK I am lazy
Does anybody have heatsink information
Make
Model
Supplier in QTY of 2 or 4 peices
I expect will be making the +/- 15V rails at 3.0 - 3.5 Amps bias
If there is several choices, I would like to see them. I would tend to use overkill when it comes to the heatsink.
Reagrds,
Aud_Mot
OK I am lazy
Does anybody have heatsink information
Make
Model
Supplier in QTY of 2 or 4 peices
I expect will be making the +/- 15V rails at 3.0 - 3.5 Amps bias
If there is several choices, I would like to see them. I would tend to use overkill when it comes to the heatsink.
Reagrds,
Aud_Mot