So I built a mini-Aleph to experiment with, and thought for use in my biamped system (fullrange 6.5" and woofers, both as dipoles on same panel, with line level active EQ and X/O).
For the circuit I mixed elements of the various iterations available. As usual after buying components I was missing some so I substituted some parts as well.
Implementation now is, one channel built, gain=20, comp caps 1.5nF, feedback R = 20k = input R, IRFP140N, +-24V from the same 375 VA 18-0-18 PSU that feeds the rest of the system, bias 1.6A, AC current gain adjusted at 100 Hz.
Now the riddles. First I listened to this one channel thrown into the system. It's biamped and the three other amps are OPA549, I used the mini Aleph for the left channel's fullrange only.
Result: sounds sweet, but strong/bloated bass. It is even more noticeable when touching the left fullrange's cone: it moves a lot more than the right's fullrange's one driven by the OPA548.
So I measured some simple parameters.
DC offset: high, 100-120 mV. Maybe my poor choice of input tail R of 330R instead of 392R or incometent matching (my best pair of 9610's, matched at 0.01V!!). OK, I accept.
Frequency range: I have a DIY fast peak meter, very precise but pot adjusted, so 1dB is the limit in resolution. Anyway I get flat over the range but -2dB down at 20kHz. Surprise but OK, I accept. (comp. cap 1.5 nF over 20k should give 80kHz cutoff no??).
Output Z: a 6.7V no load output becomes 6.62V into a 13.5 R load(that's what I had lying around...). So I calculate
Zout = R(L) * {V(0)/V(L) - 1} = 0.16 Ohms, OK.
So how come I get uncontrolled bass in a fullrange driver with Fs 80 Hz but cut off at 310 Hz by a 24 db/oct HP??.
So I measure V(out) with and without driver leads attached.
Riddle 1: now I get 2 dB down with this 8 Ohm load (no x/0, naked driver) vs. no load. I calculate output impedance at 2 Ohms, that's bad indeed.
Riddle 2: connecting or disconnecting the woofers on the same side as the mini-Aleph changes the electrical response of the mini Aleph. I get electrical outputs of +1dB at 125 Hz, +2 dB at 100 and at 60 Hz, when I connect the same side woofers, the ones on the same panel as the fullranges driven by the mini Aleph. There is no change if I connect or disconnect the other side's woofers. The bloated midbass output appears when I do connect the same side woofers.
All amps are fed from the same PSU, centrally connected, so it can't be an electrical feedback problem, else the other side woofers would have the same influence. The bloated bass only disappears when the same side woofers are disconnected so it can't be simply high Q resonance whatever.
I can only imagine that the mechanical excitation of the fullrange by the woofers operating on the same panel, creates an EMF and the mini Aleph tries to fight it by raising output. Due to the high Q driver and the high Z(out) of my mini Aleph it doesn't manage though, or even increases the resonance.
Right? Or is there something else? It sounds so outlandish to me.
Back to Riddle 1:
why is the measured output impedance so much higher with a real driver attached, than with a dummy R load? Not just much higher, but very high in the absolute? The IRFP140N should have double the transconductance of the 240's, so a single transistor pair should do as well as 2 pairs in the Aleph 3... or? And the voltage drop is the same if I disconnect everything else from that 375 VA PSU.
I could double up the output devices but I'd rather understand what's happening here first...
Sigh.
Circuit below.
For the circuit I mixed elements of the various iterations available. As usual after buying components I was missing some so I substituted some parts as well.
Implementation now is, one channel built, gain=20, comp caps 1.5nF, feedback R = 20k = input R, IRFP140N, +-24V from the same 375 VA 18-0-18 PSU that feeds the rest of the system, bias 1.6A, AC current gain adjusted at 100 Hz.
Now the riddles. First I listened to this one channel thrown into the system. It's biamped and the three other amps are OPA549, I used the mini Aleph for the left channel's fullrange only.
Result: sounds sweet, but strong/bloated bass. It is even more noticeable when touching the left fullrange's cone: it moves a lot more than the right's fullrange's one driven by the OPA548.
So I measured some simple parameters.
DC offset: high, 100-120 mV. Maybe my poor choice of input tail R of 330R instead of 392R or incometent matching (my best pair of 9610's, matched at 0.01V!!). OK, I accept.
Frequency range: I have a DIY fast peak meter, very precise but pot adjusted, so 1dB is the limit in resolution. Anyway I get flat over the range but -2dB down at 20kHz. Surprise but OK, I accept. (comp. cap 1.5 nF over 20k should give 80kHz cutoff no??).
Output Z: a 6.7V no load output becomes 6.62V into a 13.5 R load(that's what I had lying around...). So I calculate
Zout = R(L) * {V(0)/V(L) - 1} = 0.16 Ohms, OK.
So how come I get uncontrolled bass in a fullrange driver with Fs 80 Hz but cut off at 310 Hz by a 24 db/oct HP??.
So I measure V(out) with and without driver leads attached.
Riddle 1: now I get 2 dB down with this 8 Ohm load (no x/0, naked driver) vs. no load. I calculate output impedance at 2 Ohms, that's bad indeed.
Riddle 2: connecting or disconnecting the woofers on the same side as the mini-Aleph changes the electrical response of the mini Aleph. I get electrical outputs of +1dB at 125 Hz, +2 dB at 100 and at 60 Hz, when I connect the same side woofers, the ones on the same panel as the fullranges driven by the mini Aleph. There is no change if I connect or disconnect the other side's woofers. The bloated midbass output appears when I do connect the same side woofers.
All amps are fed from the same PSU, centrally connected, so it can't be an electrical feedback problem, else the other side woofers would have the same influence. The bloated bass only disappears when the same side woofers are disconnected so it can't be simply high Q resonance whatever.
I can only imagine that the mechanical excitation of the fullrange by the woofers operating on the same panel, creates an EMF and the mini Aleph tries to fight it by raising output. Due to the high Q driver and the high Z(out) of my mini Aleph it doesn't manage though, or even increases the resonance.
Right? Or is there something else? It sounds so outlandish to me.
Back to Riddle 1:
why is the measured output impedance so much higher with a real driver attached, than with a dummy R load? Not just much higher, but very high in the absolute? The IRFP140N should have double the transconductance of the 240's, so a single transistor pair should do as well as 2 pairs in the Aleph 3... or? And the voltage drop is the same if I disconnect everything else from that 375 VA PSU.
I could double up the output devices but I'd rather understand what's happening here first...
Sigh.
Circuit below.
BMK, R17/18 in your schematic are drawn incorrectly, but is that the way you have your Aleph wired-up? If so, that would explain at least part of your results.
Re-check the schematic here: http://www.diyaudio.com/forums/attachment.php?postid=15054
As you can see, R17/18 are supposed to go between the drain of the lower FET and the output. Current to the load is sensed across these, and this signal is fed back to the current source circuit.
If this isn't what's causing your problems, I can't say what it could be.
Re-check the schematic here: http://www.diyaudio.com/forums/attachment.php?postid=15054
As you can see, R17/18 are supposed to go between the drain of the lower FET and the output. Current to the load is sensed across these, and this signal is fed back to the current source circuit.
If this isn't what's causing your problems, I can't say what it could be.
I could be wrong, but...
I think that Q5 should connect directly to r19, and then r17 and r18 should be in series with the load from that point, with the feedback taken from the load side.
I could be wrong, but the simulator flips out when it is connected as you have it.
If you connect it the normal way, your offset will go up to about 2+ V, and you will need to increase your current through the front end quite a bit to compensate for it.
The bandwidth simulates -3.2db at 80Khz.
I think that Q5 should connect directly to r19, and then r17 and r18 should be in series with the load from that point, with the feedback taken from the load side.
I could be wrong, but the simulator flips out when it is connected as you have it.
If you connect it the normal way, your offset will go up to about 2+ V, and you will need to increase your current through the front end quite a bit to compensate for it.
The bandwidth simulates -3.2db at 80Khz.
Hello,
Try to rewire you circuit like the circuit diagram below.
You can even use it in balanced configuration with the Balanced
Line Stage preamplifier.
Best regards,
Kristijan Kljucaric
http://web.vip.hr/pcb-design.vip
Try to rewire you circuit like the circuit diagram below.
You can even use it in balanced configuration with the Balanced
Line Stage preamplifier.
Best regards,
Kristijan Kljucaric
http://web.vip.hr/pcb-design.vip
Attachments
MBK,
your 120mV dc offset can be cured by putting a 500ohm trimpot in place of r7 and adjusting the dc offste to 0.
From my experimentation with the circuit the 100pF cap in the feedback loop with a MUCH larger than normal NFB will twist the phase of high frequencies like there is no tomorrow, the amp will likely sound like crap. I also got very funny results playing with the 1nF compensation caps.
I don't know what a AC current gain will do but I imagine it also could cause the dc offset to misbehave and the distortion to raise.
your 120mV dc offset can be cured by putting a 500ohm trimpot in place of r7 and adjusting the dc offste to 0.
From my experimentation with the circuit the 100pF cap in the feedback loop with a MUCH larger than normal NFB will twist the phase of high frequencies like there is no tomorrow, the amp will likely sound like crap. I also got very funny results playing with the 1nF compensation caps.
I don't know what a AC current gain will do but I imagine it also could cause the dc offset to misbehave and the distortion to raise.
greyhorse said:
Greyhorse,
sorry - I drew the schematic after my notebook and after the fact. The wiring is correct, schematic was wrong. Here the correct version.
Grataku: In version 1.00 I had the normal 10k (input and feedback) and no 100 pF cap. The LF behaviour was the same. Can't comment on general sound because after the first listening and headscratching I raised the gain bec. my other amps are at 26dB, for easier swapping in and out. At that point I put the 100 pF in.
The two 1 nF (I used 1.5 nF) comp caps have been recommended elsewhere here, but I must say I had a helluva time trying to adjust AC gain, a lot of funny behavior, which i took for defective construction techniques
- anyway after a lot of fiddling and swapping all of a sudden it fell into place and I got it where I wanted it, and stable. (I adjusted my R16 until the AC over each of the .39 Ohm resistors matched). But I DO suspect especially the 1.5 nF cap on the current source to create trouble.Attachments
HiMBK said:Kristijan,
thanks, I had looked at yours too before I cooked up my circuit!
The Gate of Q7 is wired between the two 1.8K resistors.
In Kristijan's and other (for example Aleph2) the Gate is wired directly to the collector of Q6.
My 2 (Euro)cents
/Hugo - sorry Moamps, your were too fast...
Arrgh, again!
Sorry, yet again, wrong on schematic, right in the actual circuit. Corrected below, along with R10 that's actually 330R (closer to the 392 it should have been).
I should look now for the only parts that are correct in my schematics, bec. they're probably wrong in circuit haha
And for once I wanted to try these fancy circuit maker programs, to have a nice schematic, unlike all the other ones drawn by hand on a messed up piece of scrap paper. I actually tried to scan my notes but they become unreadable.
Sorry to all, didn't mean to get you guys to do my editing
Sorry, yet again, wrong on schematic, right in the actual circuit.
Corrected below, along with R10 that's actually 330R (closer to the 392 it should have been). I should look now for the only parts that are correct in my schematics, bec. they're probably wrong in circuit haha
And for once I wanted to try these fancy circuit maker programs, to have a nice schematic, unlike all the other ones drawn by hand on a messed up piece of scrap paper. I actually tried to scan my notes but they become unreadable.
Sorry to all, didn't mean to get you guys to do my editing
Attachments
grataku said:
Grataku,
well, sort of in the middle: an Aleph 3 accepting a very high dissipation per device. The voltage drop over the source resistors (0.39R) turns out at slightly over 0.6 V, for a bias of a bit more than1.5 A. So dissipation should be around 22*1.5=33W, with 50% AC current gain giving 3 A into 8 Ohms. That would require about 24 V . With 24 V rails voltage just runs out slightly before current. My current chip amps have a 3A capability (5A peak) and run from the same 24 V, this was enough power wise for the system.
So this was my reasoning - I thought the numbers added up nicely if built in this way, plus in getting the same transconductance in one IRF140 as in 2 parallel IRF240's the only drawback should be the dissipation per device. The higher bias/device should be an advantage...
So it's say an Aleph 3 with 3/4 of the bias, or a Mini-A with a bit more voltage.
Actually. Thinking that something is starving in the process and likely not current, but still the feedback is not fulfilling its purpose of lowering output impedance, I think I am getting somewhere.
I read in some thread that Mr. Pass noted that the first input transitor if left alone has a gain of 20 with the 392R resistor.
I put an 330R there so the voltage gain should be <20. Now if by virtue of my feedback choices I ask the second transistor to get up to a gain of 20, it can't actually do that: even completely closed so that the input Q gets all the current, there still will be no change since input gain is never up to 20. In other words, likely I could snip the feedback completely and it wouldn't change.
Since there is no feedback I get the full device resisitve parameters in my ouput Z: 2x transconductance + source R + output R. That can come close to the 2 Ohms that I measured! (I can't remember if I measured the resistive Zout before or after I changed the feedback to 20k).
So, back to 10k, put 392R in input tail and check. Or, put say, 511R in input tail and check
Now the other riddle with the woofer connection making the Aleph's bass output stronger ... I will test that some more by changing sides etc.
I read in some thread that Mr. Pass noted that the first input transitor if left alone has a gain of 20 with the 392R resistor.
I put an 330R there so the voltage gain should be <20. Now if by virtue of my feedback choices I ask the second transistor to get up to a gain of 20, it can't actually do that: even completely closed so that the input Q gets all the current, there still will be no change since input gain is never up to 20. In other words, likely I could snip the feedback completely and it wouldn't change.
Since there is no feedback I get the full device resisitve parameters in my ouput Z: 2x transconductance + source R + output R. That can come close to the 2 Ohms that I measured! (I can't remember if I measured the resistive Zout before or after I changed the feedback to 20k).
So, back to 10k, put 392R in input tail and check. Or, put say, 511R in input tail and check
Now the other riddle with the woofer connection making the Aleph's bass output stronger ... I will test that some more by changing sides etc.
progress update
So, I found out about the strong bass... Turns out the grounding connection was at fault. Since EQ and power amps are powered from the same PSU and tapped centrally (provisionally), some ground potential differences make trouble - problem solved by tapping the power of the power amp near the EQ and not centrally after a long line. Fascinating, a 2 dB low frequency boost in an unrelated amplifier operating in a separate band, just by wiring layout!!
For the gain / output Z and DC offset problem, swapping the 330R in the input Q tail got the DC offset to 9 mV. The output Z is lower, but still not spectacular, at about 0.75 Ohms now.
So, I found out about the strong bass... Turns out the grounding connection was at fault. Since EQ and power amps are powered from the same PSU and tapped centrally (provisionally), some ground potential differences make trouble - problem solved by tapping the power of the power amp near the EQ and not centrally after a long line. Fascinating, a 2 dB low frequency boost in an unrelated amplifier operating in a separate band, just by wiring layout!!
For the gain / output Z and DC offset problem, swapping the 330R in the input Q tail got the DC offset to 9 mV. The output Z is lower, but still not spectacular, at about 0.75 Ohms now.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.