Please elaborate. Does this mean you have lengths of wire or PCB track going from one place to another?gnd said:
mixer oscillating after recap.
- Thread starter gnd
- Start date
Help Support GroupDIY Audio Forum:
Channel ICs are originally close to each other on channel board - distance between pins is 25mm. So I made small daughterboards, 30x50mm in size, one for each channel. These boards are fitted into pin headers, and thus about 10mm above original board. Pin headers fit into original IC holes, and connect to dauhterboard. Daughterboards have new IC sockets and pcb tracks for reordering, and IC decoupling. Longest pcb track on daughterboard is maybe 30mm long, others are in range of 15mm.abbey road d enfer said:
I suspected doughterboards to cause trouble, so I removed 8 most influential channels, to get rid of oscillation. Then I added most influential back, just enough to get oscillation back, to be sure that those four ar on limit of causing oscillation. Then I removed doughterboards, soldered decoupling 100nF directly into socket and soldered sockets in place of original ICs. I placed channels back in (no pin headers, IC directly in socket), and oscillation was there again, same as it was with daughterboards. Thus I concluded, that oscillation happens even without daughterboards, or at least that daughterboards do not influence oscillation much.
....
Well, it seems you've done your homework...
Can you visualize the oscillation on the channels, or is it only on the master?
Could be that the summing amps require more feedforward compensation.
Now if channels do oscillate, I've found that EQ stages, in particular Baxendall type circuits have a tendency to oscillate. Sometimes loading the output with a Zobel (47n + 100R) cures it.
Can you visualize the oscillation on the channels, or is it only on the master?
Could be that the summing amps require more feedforward compensation.
Now if channels do oscillate, I've found that EQ stages, in particular Baxendall type circuits have a tendency to oscillate. Sometimes loading the output with a Zobel (47n + 100R) cures it.
Samuel Groner
Well-known member
Some of the opamp stages have a special feedback arrangement--resistor in series with the output, capacitive feedback from the opamp output and resistive feedback from the other resistor end. I'd be very cautious in altering the capacitor value there as it is essential for capacitive load isolation. It's possible that by plugging several boards in the capacitive load of the summing bus increases and thereby triggers instability.
Samuel
Samuel
The 5532 is not unity gain stable without a compensation cap. The compensation cap is in the 50pf to 250pf range depending on the resistance in the feedback loop.
The OP275 makes a better replacement for the TL072 as the performance is very similar, but much better. The loads that op amps are pushing inside of a most mixers is very easy for even a TL072 to drive.
Years ago I had a power amp that used a TL072 as the driver. I tried out a 5532 and the amp would self-oscillate at FULL POWER! Put the TL072 back in and it was fine. Lesson learned.
The OP275 makes a better replacement for the TL072 as the performance is very similar, but much better. The loads that op amps are pushing inside of a most mixers is very easy for even a TL072 to drive.
Years ago I had a power amp that used a TL072 as the driver. I tried out a 5532 and the amp would self-oscillate at FULL POWER! Put the TL072 back in and it was fine. Lesson learned.
? The 5534 needs a compensation cap when used with gain <3. The 5532 is unity-gain stable.toobdood said:
However, almost any opamp faced with a highly capacitance to ground on its inverting input loses stability margin. The common cure is a so-called feedforward compensation cap (across the feedback resistor). There are many cases where this cap is not needed.
It just shows that replacing an opamp with a "better" one is always risky. Any serious designer who undertakes this task starts with evaluating the stability issues. In the past, it meant measuring the open-loop gain and phase response and a good amount of math. Now, with the help of a Spice simulator, it is quite easy to pre-evaluate circuits for closed-loop stability.
toobdood said:
From the TI website
TI sez said:
Even unity gain stable opamps can be made to oscillate with lag in the NF path or poor PS stiffness.
FWIW the 5534 (the single) is only stable down to closed loop gain of about 10dB without added compensation cap.
=====
Looks like abbey beat me to the draw on this...
5532/4 is good old soldier when applies properly.
JR
Thnx for replies, guys.
In the meantime, I'm banging my head against the wall. Quite sad is my state right now....ehhh.
I suspect my problem is not a single thing, but more things combined.
Yesterday I changed four ICS back to 072, and got stability with my improvised (underpowered, original) power supply. I couldn't use it long, because it has small toroid, which gets dangerously hot after some minutes. But it run stable. Improvised PS was cca 0.5m from mixer, in empty mixer frame of second identical mixer (i combined two mixers into one, one was for spare parts...). Then I switched back to my fancy DIY PS, and oscillation was there again.
So I suspect power supply to encourage oscillations at least to some degree.
My fancy DIY PS is made like this:
1. TX: External box with toroid, rectifier diodes and 4x10000uF smoothing caps, to get DC.
2. Ground: From inbetween smoothing caps I have connected four thick ground cables cca 3m long - those go to mixer ribbon cable for all 4 separate mixer grounds.
3. Voltage: In addition I have 3 wire cable going from TX box to rectifier modules inside mixer. This cable is also 3 m long, and carries +/- 26volts, and ground for reference to rectifiers.
4. Mixer: inside mixer I have 8 boards with 317/337 rectifiers. Each rectifier board feeds 4 channels, master and monitor modules with +/- 17V.
Something must be wrong with this scheme. Today I measeured with scope between end points of ground wires. I have the following:
1. TX box 4x10000uF centre point <-> mixer rectifier boards oV reference == 100mVpp @ 1.8MHz
2. TX box 4x10000uF centre point <-> mixer ground 0A == 350mVpp @ 1.8mHz
3. TX box 4x10000uF centre point <-> other 3 grounds (0in, 0P, 0F) = cca 100-150mVpp @ 1.8MHz
I cannot succesfully tie IC power rails to grounds, because grounds themselves are dancing arround. Are ground cables too long? I fear that if I redo all channels bypassing, I will achieve nothing. I tried it with short cable to ground and 330nF capacitor on end, hoping to achieve even slight improvement, touching rails and IC legs, ribon cable contacts, other grounds, but no effect. Shouldn't it be so, that with such capacitor at one point I should hit a point, where oscillation changes, decreases or even goes away?
I think first I should move external transformer as close as possible to mixer. In my mind this would reduce groud lines resistance, and reduce oscillation in grounds. It should also help with noise, isn't it? Because with this DIY PS I have noise spectrum rising at high Freq, like hi shelf filter +20dB would be appliet to noise above 5kHz. With original underpowered PS i didn't have that.
But shouldn't remote PS scheme work fine? I know mixers have long cables for PS, Neve has 20 meters limit.... Those mixers thus also have long ground lines. Is my mistake that I have just transformer far, and rectifiers close to mixer boards? Should rectifiers be close to rectifier smoothing caps and TX?
I've seen some mixer PS schematics, and usually rectifiers are external too, in the same box with TX, rectifiers and smoothing caps. Then clean DC is going over meters of cable to mixer.
This is not practical in my case, because I have 8 boards, that would be plenty of PS cabling.
In any case, I will need to rethink grounding and power scheme, it seems. Also i will have to do something about local bypassing of all ICs, so that all rails are locally bypassed also to ground, instead of one into another, as I have it now. I need to order capacitors, 100nF multilayer ceramics, box is empty atm.
I need to do some thinking about it. Any suggestions, please shoot. Does my PS scheme seem fine, with external TX and internal multiple regulators? If yes, where do I take reference for grounds from? From remote TX box between smoothing caps? Or from internal regulator boards?
...
In the meantime, I'm banging my head against the wall. Quite sad is my state right now....ehhh.
I suspect my problem is not a single thing, but more things combined.
Yesterday I changed four ICS back to 072, and got stability with my improvised (underpowered, original) power supply. I couldn't use it long, because it has small toroid, which gets dangerously hot after some minutes. But it run stable. Improvised PS was cca 0.5m from mixer, in empty mixer frame of second identical mixer (i combined two mixers into one, one was for spare parts...). Then I switched back to my fancy DIY PS, and oscillation was there again.
So I suspect power supply to encourage oscillations at least to some degree.
My fancy DIY PS is made like this:
1. TX: External box with toroid, rectifier diodes and 4x10000uF smoothing caps, to get DC.
2. Ground: From inbetween smoothing caps I have connected four thick ground cables cca 3m long - those go to mixer ribbon cable for all 4 separate mixer grounds.
3. Voltage: In addition I have 3 wire cable going from TX box to rectifier modules inside mixer. This cable is also 3 m long, and carries +/- 26volts, and ground for reference to rectifiers.
4. Mixer: inside mixer I have 8 boards with 317/337 rectifiers. Each rectifier board feeds 4 channels, master and monitor modules with +/- 17V.
Something must be wrong with this scheme. Today I measeured with scope between end points of ground wires. I have the following:
1. TX box 4x10000uF centre point <-> mixer rectifier boards oV reference == 100mVpp @ 1.8MHz
2. TX box 4x10000uF centre point <-> mixer ground 0A == 350mVpp @ 1.8mHz
3. TX box 4x10000uF centre point <-> other 3 grounds (0in, 0P, 0F) = cca 100-150mVpp @ 1.8MHz
I cannot succesfully tie IC power rails to grounds, because grounds themselves are dancing arround. Are ground cables too long? I fear that if I redo all channels bypassing, I will achieve nothing. I tried it with short cable to ground and 330nF capacitor on end, hoping to achieve even slight improvement, touching rails and IC legs, ribon cable contacts, other grounds, but no effect. Shouldn't it be so, that with such capacitor at one point I should hit a point, where oscillation changes, decreases or even goes away?
I think first I should move external transformer as close as possible to mixer. In my mind this would reduce groud lines resistance, and reduce oscillation in grounds. It should also help with noise, isn't it? Because with this DIY PS I have noise spectrum rising at high Freq, like hi shelf filter +20dB would be appliet to noise above 5kHz. With original underpowered PS i didn't have that.
But shouldn't remote PS scheme work fine? I know mixers have long cables for PS, Neve has 20 meters limit.... Those mixers thus also have long ground lines. Is my mistake that I have just transformer far, and rectifiers close to mixer boards? Should rectifiers be close to rectifier smoothing caps and TX?
I've seen some mixer PS schematics, and usually rectifiers are external too, in the same box with TX, rectifiers and smoothing caps. Then clean DC is going over meters of cable to mixer.
This is not practical in my case, because I have 8 boards, that would be plenty of PS cabling.
In any case, I will need to rethink grounding and power scheme, it seems. Also i will have to do something about local bypassing of all ICs, so that all rails are locally bypassed also to ground, instead of one into another, as I have it now. I need to order capacitors, 100nF multilayer ceramics, box is empty atm.
I need to do some thinking about it. Any suggestions, please shoot. Does my PS scheme seem fine, with external TX and internal multiple regulators? If yes, where do I take reference for grounds from? From remote TX box between smoothing caps? Or from internal regulator boards?
...
MagnetoSound
Well-known member
Your regulators are quite a long way from the smoothing caps. Are they locally decoupled, input and output?
Divide and conquer...
If you have to, start by powering one channel with two 9V batteries, and get it stable with 5532 or whatever opamps you want to use..
Then look at the PS by it self and with a resistive load...
If it is clean connect the clean channel to the clean PS,
repeat until you have your whole system happy...
JR
If you have to, start by powering one channel with two 9V batteries, and get it stable with 5532 or whatever opamps you want to use..
Then look at the PS by it self and with a resistive load...
If it is clean connect the clean channel to the clean PS,
repeat until you have your whole system happy...
JR
abechap024
Well-known member
gnd said:
That seems like maybe a strange way to power one board...how do you have all 16 of your regulators connected??
Might be better to just use 2 regulators and 2 power transistors....
I don't know about the Studiomaster, but many british mixers have resistors right at the channels, in series with the power rails (they're used to limit the injection of PSU noise into the main ground and act like fuses in case of short-circuit). In addition to the resistance of the usual 5 meter cable, it shows that overall stability does not really depend on the rails stifness.
The regulators should have their reference connections connected as directly as possible to the central ground reference point (the point to which all channel grounds are referenced).
What happens when the mixer stars oscillating? Does it distorts audio or shuts the signal? Can you clearly identify one output where the oscillation is large and visible?
The regulators should have their reference connections connected as directly as possible to the central ground reference point (the point to which all channel grounds are referenced).
What happens when the mixer stars oscillating? Does it distorts audio or shuts the signal? Can you clearly identify one output where the oscillation is large and visible?
Exactly, 10R small resistors, just before 220uF decoupling, at each channel. They work like fuses, confirmed, burned a few...abbey road d enfer said:
When oscillating, there is no distortion, sound is the same as without oscillation, same level, same quality. Oscillation it is not specific to any output, it is the same everywhere, on grounds, on rails, on main outputs, on channel direct outputs, on aux outputs, on sends. It is not present only on inputs.
agreed... to expand upon what abbey said... The PS only needs to be stiff locally at HF to reduce tendency to oscillate if devices have marginal PSRR. A remote power supply is more like a crude supply of current with adequate voltage, but doesn't have to be strictly regulated (IMO).
I have actually done consoles with 3 terminal Vreg on each channel strip, but back in the good old days we used FP (flame proof) resistors in series with the rails, because of a very bad habit in old opamps to fail as a dead short across the supply, or for studio technicians to plug them into sockets backwards, also shorting the supply (who me?.. no it came from the factory that way. :-[ ).
The modern 3-terminal regulators are short circuit protected so eliminate the need for FP resistors, or need to replace them when they open.
JR
I have actually done consoles with 3 terminal Vreg on each channel strip, but back in the good old days we used FP (flame proof) resistors in series with the rails, because of a very bad habit in old opamps to fail as a dead short across the supply, or for studio technicians to plug them into sockets backwards, also shorting the supply (who me?.. no it came from the factory that way. :-[ ).
The modern 3-terminal regulators are short circuit protected so eliminate the need for FP resistors, or need to replace them when they open.
JR
JohnRoberts said:
John, channels are all stable by themselves.
I loaded each of 8 regulator boards with 300R resistor - it got warm. Each PS board is stable in such test.
Then I start connecting channels, in groups of four, to regulator boards.
I can add master and monitor channels, and then 5 groups of channels, in random. With any four groups all is fine, with fifth group oscillation shows up slightly. When I add sixth group, oscillation goes up. It doesn't matter which grout of channels I add last, tried different combinations in different order.
It is like I reach instability with certain number of connected channels, without any focus on some specific channels. Two channel groups are made of only 5532's, three groups of 5532 and 072, and one group only of 072. Thus it seems that if is not ICs that cause oscillations.
...
I don't have a very clear mental image of your system, but allow me to scatter shoot.
Are power supply + and - rails swinging relative to channel grounds when everything is singing?
If yes, what do you see between channel ground and these PS grounds?
If the PS ground is clean wrt channel ground, while PS + and _ are swinging up and down, the PS needs to get stiffened up wrt to it's own ground.
If the PS ground is swinging wrt the the channel ground then that path needs to get stiffened up. (easier to say than do if it is a length of wire, with inductance, etc is involved).
There is a compliance somewhere between the PS ground and channel ground that is allowing the opamps to make the rails swing up and down (same as a common emitter amplifier stage The opamp power pins are like the collectors and the opamp output is like the emitters).
FWIW you should be able to make the who thing work with 10 ohm resistors in series with every PS line and even the grounds, as long as there is adequate local PS decoupling (you need to let the current coming out of the opamp output pins, get back to the PS rails locally).
It is always logical, after you find it... been there, done that... you are not the first or last...
JR
Are power supply + and - rails swinging relative to channel grounds when everything is singing?
If yes, what do you see between channel ground and these PS grounds?
If the PS ground is clean wrt channel ground, while PS + and _ are swinging up and down, the PS needs to get stiffened up wrt to it's own ground.
If the PS ground is swinging wrt the the channel ground then that path needs to get stiffened up. (easier to say than do if it is a length of wire, with inductance, etc is involved).
There is a compliance somewhere between the PS ground and channel ground that is allowing the opamps to make the rails swing up and down (same as a common emitter amplifier stage The opamp power pins are like the collectors and the opamp output is like the emitters).
FWIW you should be able to make the who thing work with 10 ohm resistors in series with every PS line and even the grounds, as long as there is adequate local PS decoupling (you need to let the current coming out of the opamp output pins, get back to the PS rails locally).
It is always logical, after you find it... been there, done that... you are not the first or last...
JR
Samuel Groner said:
How do I test this with summing bus? It is quite possible, that summing bus is questionable, I DIYed it, and connected it instead of original one. It is interesting, that noise spectrum on outputs is not flat, but rises above some 5kHz, like Hi Shelving filter, +20dB. On channels direct outputs this is not present, noise is flat 20Hz - 20kHz.
But, removing sum bus out doesn't remove oscillation. I unplugged power from sum bus amps, osc was still there. I then physically unplugged sum bus boards, osc is still there. But oscillation changes a bit with removing of sum bus amps.
Measuring on channel direct outputs (any channel, all the same),
with sum bus amps inside, osc is 530mVpp @ 1.8MHz
with sum bus unpluged power, but boards inside, osc is 480mVpp @ 1.5MHz
and with sum bus boards physically out, osc is again 530mVpp, but at 1.35MHz
Does this mean anything?
Unplugging any group of channels (any four channels), reduces osc. With any two groups (any 8 channels), osc is gone. This works no matter sum bus amps, if they are in or out.
...
Generally the active device oscillating, will exhibit the largest voltage at it's output pin (there are hypothetical exceptions but this is a good general rule). To find the smoking gun, follow the smoke.
If concerned about bus capacitance, you can temporarily swamp it out by adding larger lead capacitance across the feedback resistor for the bus amp. (assuming unity gain stable opamps, like 5532 or TL072). If the bus capacitance is 20 pF but you have 100 pF across the bus amp it should be stable because the 100pF lead is more than canceling the 20pF lag (lag in feedback networks is a source of instability).
JR
If concerned about bus capacitance, you can temporarily swamp it out by adding larger lead capacitance across the feedback resistor for the bus amp. (assuming unity gain stable opamps, like 5532 or TL072). If the bus capacitance is 20 pF but you have 100 pF across the bus amp it should be stable because the 100pF lead is more than canceling the 20pF lag (lag in feedback networks is a source of instability).
JR
