You can always use a higher voltage cap, but not necessarily a lower voltage cap. Higher voltage electrolytics often have longer life and lower impedance, so often, higher voltage is a good idea. Check the specs in detail - for an electrolytic, you want a 105°C rating with a lifetime rating as high as possible (5000-10,000 hours), a high ripple current rating and as low of an impedance rating as possible.
The search tables of Mouser or Digikey can often help you find an ideal cap that you can actually buy, and it's much easier than poring through datasheets. Those selector tables are what I use for almost all component choices these days - saves a lot of hassle. There are datasheet parts that look great but were never shipped anywhere, or you might miss out on a superior part because you didn't scour every datasheet on the planet.
The search tables of Mouser or Digikey can often help you find an ideal cap that you can actually buy, and it's much easier than poring through datasheets. Those selector tables are what I use for almost all component choices these days - saves a lot of hassle. There are datasheet parts that look great but were never shipped anywhere, or you might miss out on a superior part because you didn't scour every datasheet on the planet.
Hi Sonic77,
Most equipment would use a 10 uF capacitor to filter noise from the zener. Keep in mind that the capacitor is in parallel with the AC impedance (which is very low).
While I have seen 220 uF capacitors in that position, they may delay the voltage buildup due to their high capacitance. You can try that to see if it is a problem or not. If it is , just stick a lower value of capacitance in.
-Chris
Most equipment would use a 10 uF capacitor to filter noise from the zener. Keep in mind that the capacitor is in parallel with the AC impedance (which is very low).
While I have seen 220 uF capacitors in that position, they may delay the voltage buildup due to their high capacitance. You can try that to see if it is a problem or not. If it is , just stick a lower value of capacitance in.
-Chris
Perhaps none? One of the 'tyrannies' of decoupling is that a bypass capacitor serves to more tightly AC couple the two nodes that the capacitor connects. This means that the potential at the 'ground' foil where the decoupling capacitor is connected will be more tightly coupled to the node at the other side of the cap. If that part of the PCB ground foil is 'quiet', then no additional junk will be coupled into the node at the other side of the bypass cap. But, symmetrically, any current noise at the non-ground node will be coupled into the ground foil at the other side of the decoupling cap.
So, a decoupling cap, if it's large enough to do anything, will potentially change the structure of signal induced currents within the PCB, for better or worse. It's really tough to say a priori how (or if) any of that will happen.
In this amplifier, it seems that the zeners serve to provide relatively low impedance voltage sources to the circuit nodes that they shunt. Their errors will be to add noise as well as drift, and a decoupling cap can help to reduce the magnitude of this zener noise.
If you want to examine this, I'd recommend adding some sufficiently large bypass caps, whose AC impedance is significantly low compared to the AC impedance of the zeners at their bias currents, and then measure the change in overall circuit noise with and without the bypasses. My gut feeling is that the zeners will be very quiet, and that noise injected into the nodes that they control will already be reduced by overall loop feedback, and thus not necessarily worth reducing by this technique.
It's also worth comparing the distortion of the overall amplifier before and after to see if you've messed things up or have improved things due to your new coupling paths caused by the bypasses.
My suspicion is that by shunting these zeners, which do little more than establish power supply voltages to circuits that have at least a bit of power supply rejection, you won't change much at all, and at best, you may reduce noise very slightly. If noise reduction is paramount, I imagine that using lower resistor values and/or 'quieter' devices in strategic places can result in much more noise improvement.
Still, is the amp noisy? Sometimes, circuits are designed really well and that any improvements to one aspect will come at the expense of something else. By all means, you're free to experiment, but it's important to be able to reliably measure the before and after state, and not just rely upon listening tests, which can be extremely unreliable.
So, a decoupling cap, if it's large enough to do anything, will potentially change the structure of signal induced currents within the PCB, for better or worse. It's really tough to say a priori how (or if) any of that will happen.
In this amplifier, it seems that the zeners serve to provide relatively low impedance voltage sources to the circuit nodes that they shunt. Their errors will be to add noise as well as drift, and a decoupling cap can help to reduce the magnitude of this zener noise.
If you want to examine this, I'd recommend adding some sufficiently large bypass caps, whose AC impedance is significantly low compared to the AC impedance of the zeners at their bias currents, and then measure the change in overall circuit noise with and without the bypasses. My gut feeling is that the zeners will be very quiet, and that noise injected into the nodes that they control will already be reduced by overall loop feedback, and thus not necessarily worth reducing by this technique.
It's also worth comparing the distortion of the overall amplifier before and after to see if you've messed things up or have improved things due to your new coupling paths caused by the bypasses.
My suspicion is that by shunting these zeners, which do little more than establish power supply voltages to circuits that have at least a bit of power supply rejection, you won't change much at all, and at best, you may reduce noise very slightly. If noise reduction is paramount, I imagine that using lower resistor values and/or 'quieter' devices in strategic places can result in much more noise improvement.
Still, is the amp noisy? Sometimes, circuits are designed really well and that any improvements to one aspect will come at the expense of something else. By all means, you're free to experiment, but it's important to be able to reliably measure the before and after state, and not just rely upon listening tests, which can be extremely unreliable.
Thanks Monty, i appreciate your feedback. I was thinking of using 100nF MKT for D1 & D2, i believe LesW said this made a big difference (although i think he used tantalum which i'm not keen on).
Anyhow i figure 100nF MKT is a low enough value that it shouldn't cause any problems, and my gut feeling is it's worth trying.
Anyhow i figure 100nF MKT is a low enough value that it shouldn't cause any problems, and my gut feeling is it's worth trying.
I'm not sure exactly which zeners are used in the Quad 909, but the mid 80s datasheets for Motorola 500mW 6.8V zeners say that they will have a dynamic impedance of around 3Ω at 5-10mA zener shunt current. So, a 100nF cap will have a -3dB point of around 500kHz in parallel with a 3Ω impedance. Basically, it isn't going to do much to noise (which should be pretty low anyway, and primarily LF) or dynamic impedance (which is already very low as well).
It very well could cause problems, since a 100nF MKT will not be physically small, and it might couple random BS into the circuit nodes that it connects to. Also, you're incrementing the "rework counter" on the amp's PC board, which is finite. So, hack away, but realize that there is little reason why a 100nF 'bypass' will do much here. It might change the coupling of currents in the PCB, but again, against a 3Ω dynamic impedance, you need a lot of capacitance to alter anything anywhere near the audio band.
Larger caps might do more, but again, against a 3Ω zener impedance, I don't see the urgent need to snub noise or reduce HF impedance. Additionally, as someone mentioned earlier, larger caps might cause startup and shutdown problems as they charge and discharge.
It very well could cause problems, since a 100nF MKT will not be physically small, and it might couple random BS into the circuit nodes that it connects to. Also, you're incrementing the "rework counter" on the amp's PC board, which is finite. So, hack away, but realize that there is little reason why a 100nF 'bypass' will do much here. It might change the coupling of currents in the PCB, but again, against a 3Ω dynamic impedance, you need a lot of capacitance to alter anything anywhere near the audio band.
Larger caps might do more, but again, against a 3Ω zener impedance, I don't see the urgent need to snub noise or reduce HF impedance. Additionally, as someone mentioned earlier, larger caps might cause startup and shutdown problems as they charge and discharge.
I am surprised that a small cap around the diode wouldn’t have been more the norm.
Aren’t diodes fast, with a higher frequency noise? Wouldn’t a .01 - .47uf film or even ceramic be best?
I have also seen the 10uf commonly used as well around a diode used to drop voltage to an op amp supply.
Aren’t diodes fast, with a higher frequency noise? Wouldn’t a .01 - .47uf film or even ceramic be best?
I have also seen the 10uf commonly used as well around a diode used to drop voltage to an op amp supply.
These diodes aren't switching - they're essentially zener voltage references, continuously biased, so there's no HF switching noise to be removed. Further, their dynamic impedance, at their breakdown current bias point, is on the order of 3 ohms, so if you expect a cap to have any effect whatsoever on the zener noise (which does exist), its impedance must be at least on the order of 3Ω, and a 0.47µF cap is never going to be able to provide an impedance low enough to make any difference over the audio band, the only place where this diode noise is going to be important.
Mind you, the Quad 909 circuit is not "normal", but it is a well designed circuit. Specifically, these zener diodes are absolutely optimal - no other zener breakdown voltage values will work better in terms of having lower dynamic impedance, and given that, no small capacitor will reduce their noise, which isn't actually a problem. I can authoritatively state that these zener bias diodes do not have switching "gunk" that needs to be bypassed, and even if you want to reduce their noise, anything as small as 470nF is not gonna make any difference at all into their 3Ω dynamic impedance.
Edit:
I just did some thumbnail calculations, and a 10µF cap will have an impedance of ~800Ω at 20kHz. So, it will have almost unmeasurable effect event at the top of the audio band, and even less at lower frequencies. A 10,000µF cap will have an effect, as it will have a 20kHz impedance of 0.8Ω but it's gonna be pretty huge, and act as a "coupling agent" into the circuit. Furthermore, IIRC, the node that this zener defines has some power supply rejection built into it, so it's not a "dangerous" node anyway.
Mind you, the Quad 909 circuit is not "normal", but it is a well designed circuit. Specifically, these zener diodes are absolutely optimal - no other zener breakdown voltage values will work better in terms of having lower dynamic impedance, and given that, no small capacitor will reduce their noise, which isn't actually a problem. I can authoritatively state that these zener bias diodes do not have switching "gunk" that needs to be bypassed, and even if you want to reduce their noise, anything as small as 470nF is not gonna make any difference at all into their 3Ω dynamic impedance.
Edit:
I just did some thumbnail calculations, and a 10µF cap will have an impedance of ~800Ω at 20kHz. So, it will have almost unmeasurable effect event at the top of the audio band, and even less at lower frequencies. A 10,000µF cap will have an effect, as it will have a 20kHz impedance of 0.8Ω but it's gonna be pretty huge, and act as a "coupling agent" into the circuit. Furthermore, IIRC, the node that this zener defines has some power supply rejection built into it, so it's not a "dangerous" node anyway.
Last edited:
R1 and R2 are not in the feedback loop - they're ahead of the amp. There are a few other resistors that really don't handle signal, so they could be considered "out of the loop": the zener bias resistors R41 and R3. The resistor R33 (the output RC termination resistor) isn't in the loop, but it is in the signal path, as is R1 and R2, and I guess that's about it. However, all of these parts do have an effect, and some resistors have more effect than others.
I'm not sure where any improvement can be made in this amplifier, especially since it is such an unusual topology - standard rules may not apply. The circuits I'm personally working on now depend upon the PC board foil geometry, and so what can be observed from a schematic is somewhat incomplete - the complete implementation on a PCB is sometimes very important.
Since the 909 is not a haphazard design, I would think that it is probably on the side of "mostly correct", but still, sometimes cost limits prevent completely optimal parts from being used. Still, the normal 'trick' of matching parts to make an amp more symmetric might not matter in this circuit - it really is an unusual circuit, so the usual rules really ought not apply. I'd tread lightly with this circuit, and maybe try to just enjoy it? Or, slather money on it by "blueprinting" it with expensive, close tolerance components? If I come up with any brilliant insights, I'll let you know, but this topology is definitely not "normal", so generic modification principles should be viewed with caution.
If you're adventurous, try simulating this with LTspice. In this way, you can re-wire the circuit endlessly and examine the distortion with extreme resolution - over 200dB if you set the parameters right. LTspice is my first 'go-to' when designing a circuit, as it at least does the tedious calculations that put the design into the right ballpark, and then allows me to tweak it until it seems right, at least on paper. Then, a physical construction will show me what the models do not show me, and the design can then be finished with less hassle.
Best of luck!
Thankyou everyone, and monty especially, thankyou.
There are definitely things that improve the sound of 909, i have heard of numerous people changing the input coupling cap (C2) from the stock MKT to Polypropylene with great results (much more clear sound), and numerous people have reported big differences form replacing C7 on the boards. These are tested and proven upgrades.
I'm looking for icing on the cake, and though maybe it wouldn't hurt to replace some key resistors with Vishay RN60. I do hear hear what your saying however about the 909 not being a standard topology, so no guarantee replacing any resistors will help, but it doesn't hurt to try.
I would like to replace R1 and R2, can anyone else recommend are there any other resistors of significance that might be worth changing while i'm at it?
Cheers : )
There are definitely things that improve the sound of 909, i have heard of numerous people changing the input coupling cap (C2) from the stock MKT to Polypropylene with great results (much more clear sound), and numerous people have reported big differences form replacing C7 on the boards. These are tested and proven upgrades.
I'm looking for icing on the cake, and though maybe it wouldn't hurt to replace some key resistors with Vishay RN60. I do hear hear what your saying however about the 909 not being a standard topology, so no guarantee replacing any resistors will help, but it doesn't hurt to try.
I would like to replace R1 and R2, can anyone else recommend are there any other resistors of significance that might be worth changing while i'm at it?
Cheers : )
Last edited:
The Quad 909 model is Chinese built and uses "Capxon" brand electrolytics, I believe. These are known to be short-lived and bulging is obviously a sign of failure. According to the Dada site, Kendeil or Kemet/BHC capacitors are preferred but any reputable brand electrolytic suited to audio applications will be fine. You should look for the obvious specs that show similar or slightly better capacitance, voltage rating, ripple current and temperature ratings and naturally, one that fits the space comfortably and has the correct style of connector pins and spacing.
Here is just one source with examples to choose from - Kendeil: Consumer Electronics | eBay
Here is just one source with examples to choose from - Kendeil: Consumer Electronics | eBay
Last edited:
Follow up questions
Couple of questions for you:
1) Did you use the 606 clone boards on ebay (the latest 606 board and the 909 board)?
2) What were your substitutes for unavailable components particularly semiconductors?
3) Did you build the opamp based balanced switched inputs, colloguially the AmpBus? If so, how did you implement these?
Couple of questions for you:
1) Did you use the 606 clone boards on ebay (the latest 606 board and the 909 board)?
2) What were your substitutes for unavailable components particularly semiconductors?
3) Did you build the opamp based balanced switched inputs, colloguially the AmpBus? If so, how did you implement these?
Just got a very nice 909. Besides cap replacement and many other mods suggested on this thread (which I plan to check into), I've also removed the Quadlink board and connected the inputs directly to the audio boards as on the 405, which seems to improve sound by much. However the LED light is lost now as its power comes from the Quadlink board... Anyone has a suggestion as of how the LED can be activated otherwise?
Last edited: