Hello,
I hope I'm not being too redundant here, but when powering a preamp (actually a stepped attenuator followed by 2 active xovers using opa2134 or LM4562 ICs) with batteries, +/-12v from SLAs, should you put filter caps between the + and - rails to com, or any bypass, snubber or whatever caps to lower ESR and provide maximized energy transfer. I have bypass caps close to the IC's supply pins. Also, can electrolytics be of too large capacity? And I gather the use of regulators is not recommended? Also, what values, types of caps would be best suited for the purposes intended. All this can be rather confusing. After reading several threads here it almost seems that no psu is up to true 'audiophile' performance!
I hope I'm not being too redundant here, but when powering a preamp (actually a stepped attenuator followed by 2 active xovers using opa2134 or LM4562 ICs) with batteries, +/-12v from SLAs, should you put filter caps between the + and - rails to com, or any bypass, snubber or whatever caps to lower ESR and provide maximized energy transfer. I have bypass caps close to the IC's supply pins. Also, can electrolytics be of too large capacity? And I gather the use of regulators is not recommended? Also, what values, types of caps would be best suited for the purposes intended. All this can be rather confusing. After reading several threads here it almost seems that no psu is up to true 'audiophile' performance!
You should always add decoupling close to the opamp itself. Even if the batteries were a perfect DC voltage source of zero impedance, by the time you add wiring/print resistance its not quite as perfect when it gets to the opamp. Any rail electrolytics need only be small, say 100uf. In practice an SLA is as good as it gets.
(And please... use suitable fuses when dealing with SLA's. The current capability is in the hundreds of amps under fault conditions)
(And please... use suitable fuses when dealing with SLA's. The current capability is in the hundreds of amps under fault conditions)
If you use a dual polarity supply for your opamp then the PSU zero volts line will pass load current.
The PSU zero volt line must be included in the decoupling regime that you implement.
That requires a +ve to 0v decoupling cap at the opamp and needs a -ve to 0v decoupling cap at the opamp.
The loop area for the opamp to decoupling cap and back to the opamp MUST be kept small and short. If not the decoupling can't work.
Mooly said
The PSU zero volt line must be included in the decoupling regime that you implement.
That requires a +ve to 0v decoupling cap at the opamp and needs a -ve to 0v decoupling cap at the opamp.
The loop area for the opamp to decoupling cap and back to the opamp MUST be kept small and short. If not the decoupling can't work.
Mooly said
So all I need are a couple of 100uf caps (any problems with larger values?) between the +/- rails and com and bypassing the +/- IC pins(see attached, how I did it. Is this adequate? I've seen other posts recommending the .1uf ceramic + a 1uf polypropylene, maybe a mica or polystyrene, changing the 10uf electro to a PPL and somewhere adding a couple of chokes, diodes and somehow getting the kitchen sink in there to). And you mentioned fuses, what would recommend? There'd be 2 12v 5Ah SLA batts powering 2 opamps.
Attachments
There's no real problem going bigger with the caps but I would question whether you need to. Battery supplies are super clean and so all you really need to do is maintain a low AC impedance at the opamp and that generally means small caps.
Adding chokes will raise the AC impedance of the supply and make it more susceptible to pick up stray fields and interference. Batteries are totally different to AC supplies.
Fuses should be fitted close to the battery and could be as low as 100ma rated (as long as you don't fit massive caps). If the preamp is a permanent design on a proper PCB then fitting ICP's (integrated circuit protectors) is a neat idea. These are fast acting fuses in a small 2 legged package that are solderable.
Adding chokes will raise the AC impedance of the supply and make it more susceptible to pick up stray fields and interference. Batteries are totally different to AC supplies.
Fuses should be fitted close to the battery and could be as low as 100ma rated (as long as you don't fit massive caps). If the preamp is a permanent design on a proper PCB then fitting ICP's (integrated circuit protectors) is a neat idea. These are fast acting fuses in a small 2 legged package that are solderable.
Your pic shows two types of supply rail decoupling.
The HF decoupling is a 0.1uF (=100nF) is a cer, that is a ceramic type. Use x7r.
This gets soldered very close to the Power pins and connects to power ground via a VERY SHORT route. Long traces are no good for HF decoupling.
The Electrolytic are the MF decoupling. These can be connected with slightly longer traces to the power gound. Try to achieve less than 10mm long on each side of the capacitors.
Do not fit ceramic, or any low esr, capacitor across the pins of an electrolytic.
A battery doesn't change any requirements for good circuit bypassing. active devices need bypass caps dictated by the amount of gain and loading (see data sheet ) and PCBs or breadboards need board level bypassing dictated by what's on them and how far the battery or power supply is from it.
start with a good linear lab supply or modern LDO Vregulor IC hard to go wrong.
PCB board level bypass for what yer building 200-470uF will be fine.
hehe a case of forum noise exceeding the desired result / signal I'm afraid.
start with a good linear lab supply or modern LDO Vregulor IC hard to go wrong.
PCB board level bypass for what yer building 200-470uF will be fine.
Last edited:
I had a couple of 4800uf CG caps which I used for the battery supply (see fig 1). The star gnd point is the same as with the regulated psu, which has been unplugged and. except for that gnd, is out of the circuit. The batt psu works very well except for one thing. I get a very significant hum which varies with the volume setting. The input comes from my pc and it makes no difference if it's plugged in or not. When I switch the batteries off, the hum goes away. I can't say my grounding scheme is ideal, but this problem wasn't there with the AC psu. I fed a sine wave over the hum and determined that it is primarily 180hz, with some 60hz thrown in. So the big question is, of course, why am I getting this hum with batteries?
Happy Holidays
Happy Holidays
Attachments
probably a ground loop between the PC to your testing apparatus.
you need more trouble shooting> batterys off means you cant hear it but the loop remains.
count how many piece of gear then look at the way stuff is plugged in.
you need more trouble shooting> batterys off means you cant hear it but the loop remains.
count how many piece of gear then look at the way stuff is plugged in.
Last edited:
Hi Andrew,
Is it advised to choose a reservoir cap with a better ESR and inductance than the SLA before the local decoupling of the opamp to increase the performance of the powersupply?
For instance (as it's a pre to avoid ceramic): a 5mm pitch (for a not too bad inductance) Panasonic FC > to 3000 uF (for the lowest ESR possible from an electrolytcap) : I mean with the goal to putt it the closest possible to the opamp (swapping the 100 uF). Or is it not working as the current needs to comeback to the zero volt towards the batteries center tap?
(Of course there are also some polymer around> 560 uF with even lower ESR and smaller inductance but my experience says me they are only suited for digital circuits !)
and near the opamp voltage pins : Wima MKT 2.5 mm pitch, between 0.1 to 1 uF ? (one lead of the cap on the opamp voltage pin to have the shortest ground loop possible) ?
The electrolytic near the battery acts in conjunction with the battery's source impedance to create a filter.
The esr of the electrolytic will reduce the performance of the filter slightly. You will struggle to measure that loss of performance. I don't think it would be audible if the local decoupling is properly implemented.
Spending money on exotic capacitors will rarely if ever improve performance.
After the current has passed along the cables, the client circuit sees those cables as increased source impedance. Any capacitance after those cables creates a further filter.
Local deoupling has two performance enhancing attributes.
a.) filtering the unwanted signals on the supply lines.
b.) supplying the power to operate the client circuit where the inductive impedance of the cables and traces is high enough to strangle the current supply. A glitch (transient voltage drop when current increases rapidly) is the symptom of high source impedance in the supplt rail/cable/trace.
This latter attribute demands that the HF decoupling be at the supply to load return pins. But most opamps and some amplifiers don't have a load return pin. You need to create that route for the load return current. That route MUST be VERY short and compact. You do not take the load return route some inches away to a higher impedance PSU Zero Volts point/trace/plane. It MUST be local to the chip.
The esr of the electrolytic will reduce the performance of the filter slightly. You will struggle to measure that loss of performance. I don't think it would be audible if the local decoupling is properly implemented.
Spending money on exotic capacitors will rarely if ever improve performance.
After the current has passed along the cables, the client circuit sees those cables as increased source impedance. Any capacitance after those cables creates a further filter.
Local deoupling has two performance enhancing attributes.
a.) filtering the unwanted signals on the supply lines.
b.) supplying the power to operate the client circuit where the inductive impedance of the cables and traces is high enough to strangle the current supply. A glitch (transient voltage drop when current increases rapidly) is the symptom of high source impedance in the supplt rail/cable/trace.
This latter attribute demands that the HF decoupling be at the supply to load return pins. But most opamps and some amplifiers don't have a load return pin. You need to create that route for the load return current. That route MUST be VERY short and compact. You do not take the load return route some inches away to a higher impedance PSU Zero Volts point/trace/plane. It MUST be local to the chip.
one short lead and one long lead does no good at all. It's the TOTAL route length and the TOTAL impedance of the flow and return ROUTE/LOOP that determines the loop impedance. It's the loop impedance that determines the capability of the local decoupling to supply transient current when current demand suddenly changes.
Last edited:
Thank you so much Andrew,
I appreciate a lot the translation of each technical word in its usefull local implementation in the layout and further desciption on the effects (glitch, etc) : very didactic, it helps a lot 🙂.
I appreciate a lot the translation of each technical word in its usefull local implementation in the layout and further desciption on the effects (glitch, etc) : very didactic, it helps a lot 🙂.
Not sure if this is relevant, but I've always liked to have diodes appropriate for the situation on the wires from the battery bank to the amp, in my mind everything that is connected can introduce noise at any point in the chain. Never had any problems with noise when using battery supply, but then again the largest I've ever had was battery banks for dual mono 48V (+/-24V) so 4*12V per channel. Really miss those, but sadly not compatible with small children.
Used Elliott's circuit for reducing ground noise (resistor+cap+rectifier), then connected that to the battery bank, and also a point from that star to the amp. Star is good, follow the star! 🙂
Used Elliott's circuit for reducing ground noise (resistor+cap+rectifier), then connected that to the battery bank, and also a point from that star to the amp. Star is good, follow the star! 🙂
I assume LiFePo4 used in digital devices these days to be good enough for a pre !
A123 brand have some 3.2 V/3 Ah with as low as 6 Mohms ESR, so you can putt some in serie with less foot print maybe and they will have a better ESR than a SLA...
A123 brand have some 3.2 V/3 Ah with as low as 6 Mohms ESR, so you can putt some in serie with less foot print maybe and they will have a better ESR than a SLA...
By the time you add the required protection circuitry it is a bit higher in my tests. But it is still lower than a 7 Ah SLA.
If you want to add a parallel capacitor to either you will need a minimum of 200,000 uF to have any effect. Batteries damp noise from the circuit down to DC. Capacitors do not. Both are limited by power supply conductor resistance.
So for a bypass at the circuit devices 10 uF is more than enough at the device rail to common. This can have a .1 uF ceramic surface mount even closer to the device. I would use NPO as they are larger, easier to solder and not piezo electric. Naturally the cost way more than X7R.
Finally at any place current is drawn place a third capacitor between both supply rails. In power amplifiers I use 300 uF film capacitors for this. For a preamp 10 uF should do. But I would try to keep it a film capacitor.
Very interesting, but I have a couple of questions. 1- Where do you put the parallel 200kuf cap and what is it supposed to do. 2- "Finally at any place current is drawn place a third capacitor between both supply rails" do you mean between the + and - rails (not with com) and is placement critical? (Does all this have to do with lowering ESR?) And finally, what and where are the load return pins?
By the way, that hum problem turned out to be because I inadvertantly left an input plugged in my tv. SQ with the batt supply has significantly better resolution, dynamics, imaging and immediacy. You seem to 'hear more music'.
By the way, that hum problem turned out to be because I inadvertantly left an input plugged in my tv. SQ with the batt supply has significantly better resolution, dynamics, imaging and immediacy. You seem to 'hear more music'.
I would not recommend a 200kuF (200mF) capacitor as the filter after a battery supply.
a.) it leaks more than a 2mF capacitor and "wastes" the battery current/charge.
b.) it needs to be charged at every switch ON and that "wastes" battery charge.
c.) that level of filtering of battery noise and cable interference is not needed.
F-3dB with a source impedance of 10mOhms would be 80Hz for 200mF and 7kHz for 2m2F
a.) it leaks more than a 2mF capacitor and "wastes" the battery current/charge.
b.) it needs to be charged at every switch ON and that "wastes" battery charge.
c.) that level of filtering of battery noise and cable interference is not needed.
F-3dB with a source impedance of 10mOhms would be 80Hz for 200mF and 7kHz for 2m2F
A battery filters down to DC. Even if the battery has .1 ohm resistance what value capacitor do you need to equal that? Answer infinite.
The definition of capacitance is C=Q/V. The important part of the first derivative is C x dV/dT = i. Now most folks get that F = 1/T. So they assume that if the lowest audio frequency is 20 hertz then dT = .05. As this is a power supply not a series pass use that is too simplistic. The power supply filter capacitor may be recharged 120 times a second but the is an effective series resistance in any power source before the filter capacitor. So instead we need to look at the envelope of the signal. IE when a drummer strikes the drum a surge of power is required for as much as 2 seconds as the note decays. So you would want your power supply not to sag a noticeable amount during that time. Not being noticeable would be around 3% or less. Just to make things a bit tougher on the filter capacitors, maximum low frequency energy in music is around 150 hertz.
So the beat frequency of 30 hertz will show up at some level. As a minimum that should be 30 dB down.
Now adding capacitors to add surge filtering to a battery power supply requires stupidly large capacitors to have any effect. However with virtually all cases the measurable effect will not have any audiable change.
So the only challenge is to make sure your power distribution traces or wiring doesn't screw things up.
As to where to place a rail to rail capacitor, if you are using an op-amp as the output, right across the + & - power pins. This is in addition to the normal capacitors to ground.
Conclusion is that with batteries you are not trying to get rid of power supply noise, you are trying to damp the circuit load noise. This is best done at the noise source. That is where the circuit connects to the rails.
The definition of capacitance is C=Q/V. The important part of the first derivative is C x dV/dT = i. Now most folks get that F = 1/T. So they assume that if the lowest audio frequency is 20 hertz then dT = .05. As this is a power supply not a series pass use that is too simplistic. The power supply filter capacitor may be recharged 120 times a second but the is an effective series resistance in any power source before the filter capacitor. So instead we need to look at the envelope of the signal. IE when a drummer strikes the drum a surge of power is required for as much as 2 seconds as the note decays. So you would want your power supply not to sag a noticeable amount during that time. Not being noticeable would be around 3% or less. Just to make things a bit tougher on the filter capacitors, maximum low frequency energy in music is around 150 hertz.
So the beat frequency of 30 hertz will show up at some level. As a minimum that should be 30 dB down.
Now adding capacitors to add surge filtering to a battery power supply requires stupidly large capacitors to have any effect. However with virtually all cases the measurable effect will not have any audiable change.
So the only challenge is to make sure your power distribution traces or wiring doesn't screw things up.
As to where to place a rail to rail capacitor, if you are using an op-amp as the output, right across the + & - power pins. This is in addition to the normal capacitors to ground.
Conclusion is that with batteries you are not trying to get rid of power supply noise, you are trying to damp the circuit load noise. This is best done at the noise source. That is where the circuit connects to the rails.
- Status
- Not open for further replies.