True. But I wonder how the leakage currents effect audio quality. Most are well within safety requirements, and I have not found any relationship with audio quality yet.
Rating of power supplies seem to be thermal related, I don't really know whether there is a voltage related spec or not. While it does take some effort to get a SMPS to work well with audio circuits, I think the same effort is required for linear supplies as well. I like SMPS for the size, and it's also easier to control the weight balance of the whole device.
Additional caps are not the bigger the better. The higher the frequency, the smaller the caps is probably a good way to put it.
Talking with lots of people, the general impression seems to be SMPS makes the sound more lean and a bit harsh. I think this may be the case when power and audio amplification are considered as two individual devices. To improve the situation, it seems that power supply and audio amplification needs to be considered an integrated device.
I heard it ! For a few minutes I heard the board I thought dead play music.
Since I have new 3886, 318 and BAV99 on the way, I re-flowed the solder on D1 & D2 and on the op-amp. Hooked it up and it played. My jaw dropped open and then formed a big grin. I fetched my board that was working before, powered down and hooked it up. WHen I powered up again, it (the good board) did not work. I tried swaping speaker leads and had the wire pull out of the connector. I thought, could all my problems be a bad contact on the crimped spade connector ? I then soldered the speaker wires to the board thinking this must be it. I powered up and ...... neither board worked.
When I power up, I think I hear two clicks. What is the function of the BAV99 on the FE ? What conditions trigger the mute pin on the FE ? Did I understand KSTR right, hooking the speaker ground lead to the Ground spade between the PS caps bypasses the relay ?
Since I have new 3886, 318 and BAV99 on the way, I re-flowed the solder on D1 & D2 and on the op-amp. Hooked it up and it played. My jaw dropped open and then formed a big grin. I fetched my board that was working before, powered down and hooked it up. WHen I powered up again, it (the good board) did not work. I tried swaping speaker leads and had the wire pull out of the connector. I thought, could all my problems be a bad contact on the crimped spade connector ? I then soldered the speaker wires to the board thinking this must be it. I powered up and ...... neither board worked.
When I power up, I think I hear two clicks. What is the function of the BAV99 on the FE ? What conditions trigger the mute pin on the FE ? Did I understand KSTR right, hooking the speaker ground lead to the Ground spade between the PS caps bypasses the relay ?
it limits LM318's output to circa 28V.
The amp works perfectly with the two BAV99 and 27V zener (D3) removed, it's a safety network.
None, it is always off (mute goes to V- via R2)
Yes, it bypasses all the DC protection circuitry.
But don't hook speakers before measuring DC offset between OUT and PGND (must be less than 30mV)
Sorry to hear you're out of luck.
LM3386's mute pin functionality is disabled on the FE.
Attaching speaker return lead to PGND overrides the relay. Not recommended.
Use a dummy load (8R2, 1Watt) first, when you seeing it go up in smoke you know why.
Short the input on first power up (with a shorted RCA plug).
And check the supply voltages (at the LM318), offsets etc with a multimeter.
The two BAV99+27V zener provide a clipping function when the ouput voltage reaches a certain level. Normally it is not active (except when you happen to have the Zener installed the wrong way, then the first output pulse above 16V or so (pos or neg) might easily kill the LM318, the LM3886 will pull the LM318 output beyond the +-14V rails, no good).
LM3386's mute pin functionality is disabled on the FE.
Attaching speaker return lead to PGND overrides the relay. Not recommended.
Use a dummy load (8R2, 1Watt) first, when you seeing it go up in smoke you know why.
Short the input on first power up (with a shorted RCA plug).
And check the supply voltages (at the LM318), offsets etc with a multimeter.
I have, and can be quite strong indeed. Any leakage current between amp, amp channels, and source will drop voltage on the interconnect shields (and the 1R in case of the FE) which will bleed directly through into the audio.
I strongly suspect (also because I can easily measure it) that many many heard system effects of different interconnects, mains cables, swapped mains plug polarities, mains filters etc is boiling down to different amounts and quality of leakage currents and the effect they raise.
A very typical case is a notebook powered from mains (with SMPS which have said high leakage current potential to earth) connected to an amp or active speaker with earth-grounded audio GND (with or without "isolation" resistor) and unbalanced interconnects, using a USB external audio interface. This setup pretty much always makes audbile hum/buzz. Run the notebook off it's battery and all is well.
Last edited:
I have - the leakage capacitance means that the switching currents from mains want to find their way back to ground. In a system interconnected with single ended cabling (coax and RCAs) the leakage currents flow in the screens and therefore appear in series with the wanted audio signal.
This article outlines many of the issues - a 1MHz 4.5uA current source forcing interference onto grounds looks to me to have an impact on audio performance. At 1MHz even a low 1uH inductance (by no means atypical for an interconnect) has 6.3johms of impedance giving roughly 30uVRMS of aggressor signal. And this is a measly LED SMPSU - scale up perhaps 10-100X for a power amplifier PSU.
Power Tip 40: Common-mode currents and EMI in non-isolated power supplies
Interconnect impedance match with the input load is what I have discovered as the main effect on sound quality, this was also addressed in the Sumiko OCOS report. This had been confirmed using a CD player that had both balanced and unbalanced output to amplifier with both inputs. The results can either be indistinguishable or better with the impedance properly matched, depending one other features of the interconnect. Comparison was made between constant impedance unbalanced interconnect and balanced interconnect between the same equipment.
Less leakage does not seem to be directly related to audio quality. At least during design of mains power filter did not show this relationship for me. I have also not found any noticeable difference that can be allocated to Class I device connected to Class II device or connected with the ground pin floating. Most of the time I listen from a notebook with the PS plugged in, since my wall sockets are only Class II, I never had any problems. I did not like using USB cards since Win7 because mine never played without breaks. Using express card was better for me, and never created noise problems. Messing digital ground with analog ground is not good either, which is probably hard to avoid using USB soundcards.
I have experienced large leakage designs that sound good, even at levels higher than the safety standard. I did get a few shocks.
Less leakage does not seem to be directly related to audio quality. At least during design of mains power filter did not show this relationship for me. I have also not found any noticeable difference that can be allocated to Class I device connected to Class II device or connected with the ground pin floating. Most of the time I listen from a notebook with the PS plugged in, since my wall sockets are only Class II, I never had any problems. I did not like using USB cards since Win7 because mine never played without breaks. Using express card was better for me, and never created noise problems. Messing digital ground with analog ground is not good either, which is probably hard to avoid using USB soundcards.
I have experienced large leakage designs that sound good, even at levels higher than the safety standard. I did get a few shocks.
Last edited:
This article does not address audibility issues. I have worked on mains filter for a SMPS, there is no discovered relations between leakage current and amplifier sound quality. There needs to be very specific match between SMPS and mains, and SMPS and circuit it supplies.
Pretty much the same as the Sumiko OCOS report: Interconnect impedance matching the equipment input impedance. Since I believe in constant load impedance over the audio range (this may change because I am still investigating effects beyond audio range), my interconnects are designed to be as such.
During one test configuration, I could not hear a difference, but the other auditor expressed his impression, but added that his day to day mood changes would have more influence than the difference he could hear.
The later design provided less sound coloration that did not need quick switching as was done the first time.
Last edited:
At 5kHz++ (typical) a coax interconnect will be a good common-mode choke (the whole deal behind using coaxes at HF) so the shield drop is not a problem (but HF ingress is, when the shields do not terminate directly onto the metal chassis).
The big problem in many SMPSU's is the ubiquituous "EMC cap" accross HF xformer pri and sec (1nF and more). This couples the FULL mains voltage plus any ripple (LF and HF) at the main supply cap to the secondary (our GND) via this EMC capacitor. There are ways around this but the usual SMPS makers (using standard application notes) are not aware of it.
The first thing I do when a SMPS comes into the lab is hook a AC voltmeter (with 10Meg input impedance) between output and earth ground, typically 100V can be seen (at 230V mains) regardless of mains plug polarity. More detail can be seen (also in the real world setup) with a good current clamp on the interconnect or whatever cable I'm inspecting.
The big problem in many SMPSU's is the ubiquituous "EMC cap" accross HF xformer pri and sec (1nF and more). This couples the FULL mains voltage plus any ripple (LF and HF) at the main supply cap to the secondary (our GND) via this EMC capacitor. There are ways around this but the usual SMPS makers (using standard application notes) are not aware of it.
The first thing I do when a SMPS comes into the lab is hook a AC voltmeter (with 10Meg input impedance) between output and earth ground, typically 100V can be seen (at 230V mains) regardless of mains plug polarity. More detail can be seen (also in the real world setup) with a good current clamp on the interconnect or whatever cable I'm inspecting.
Last edited:
True enough - to get a handle on those I suggest looking out for a paper (which I've not read as its behind the AES paywall) of Deane Jensen - google for 'Spectral Contamination'. Bill Whitlock carries on the good work over at Jensen Transformers nowadays so he might be a good person to ask about audibility.
It very much depends on the whole system grounding - whether balanced or unbalanced cabling is used for example. The mechanisms for coupling are clear though - much depends on the resistance of amp to RF ingress. This I believe is the reason many audiophiles turn up their noses at IC opamps
What are your findings? Please give a summary of what impedances you have on your interconnects and what impedances at the Source and at the Receiver give good or better sound results.
What?
You're saying that an interconnect center conductor should have, say, 47kOhm//100pF impedance to provide a tricky high impedance 1:1 voltage divider?
My RF/EMC engineering is a bit rusty but using coaxs for HF assumes them being used in normal mode (where the return currents for the centre conductor return on the inside of the shield) not common-mode. In the CM case as here, don't the currents travel on the outside of the shield and therefore transfer impedance is important?
Have you ever seen an RF-grade shield termination being used on an audio amp?
I forgot to mention. I took the FE's off my one large heat sink and attached them to separate computer heat sinks. Even though only one channel was working both heat sinks got warm. Would this suggest your first guess is correct; that I have a (almost) dead LM318 ?
DC Volt with the meter set on 200m V --- is 1.8 for the working side and 0.8 for the non-working side.
BTW, I fixed the dead display segment on my meter and gave it a new battery; it is doing much better.
DC Volt with the meter set on 200m V --- is 1.8 for the working side and 0.8 for the non-working side.
If that is what I did wrong, would the zener also be killed or could it be reused ?
I'm truly a novice here. Is this done on the pads with the LM318 removed or can it be done with the op-amp in place ? Which LM318 pins provide the supply voltage ?
BTW, I fixed the dead display segment on my meter and gave it a new battery; it is doing much better.
I am not sure how difficult it is to visualize exact impedance curve shape match. However you want to implement it. Generally, the lower the impedance on the receiving end is less influenced by interconnect impedance variation.
This is not even the right thread to discuss this.
Last edited:
No. Only the impedance shape has to match. Sorry that I was not clear.What?
Yes, the HF leakage currents travel on the outside of the shield (that's why I use -- and recommend -- non-isolated RCA jacks connecting directly to the chassis, HF-styleMy RF/EMC engineering is a bit rusty but using coaxs for HF assumes them being used in normal mode (where the return currents for the centre conductor return on the inside of the shield) not common-mode. In the CM case as here, don't the currents travel on the outside of the shield and therefore transfer impedance is important?
Have you ever seen an RF-grade shield termination being used on an audio amp?
). The HF/MF (and upper audio range) return currents run on the inside. LF and DC goes along the path of lowest resistance (if there is a second path, that is. And then shield resistance vs. other paths will determine how high the voltage drop along the shield is).These are complicated matters not easily analysed in the real world.
Henry W. Ott's legendary puplication "Noise Reduction in Electronic Systems" contains a wealth of information which I found very helpful (I'm not an HF expert, too, in fact many things are still a complete mystery to me)
Last edited:
Hmmm... I still cannot imagine the exact scheme, what's an "impedance shape"?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.