Thanks Toni. Actually, I had found that shifting it to the left of R1 made routing a lot easier. Here's a snip of the bottom layer of the board (right channel) showing the relay K1.
(I need to add + pin identification on the bottom of the board for the pol caps - a tip I learnt from Mark.)
(I need to add + pin identification on the bottom of the board for the pol caps - a tip I learnt from Mark.)
Attachments
Last edited:
View of ground planes. I set Width=10mil and Isolate=24mil. 24mil looks like a lot but this calculator tosses or 25.4mil for 12V...
Images of latest circuit and full board (ground planes not displayed) attached also. (Board remains 120 x 50mm.)
I will double-check the connector footprints and everything else tomorrow.
Thanks for the help guys!
Images of latest circuit and full board (ground planes not displayed) attached also. (Board remains 120 x 50mm.)
I will double-check the connector footprints and everything else tomorrow.
Thanks for the help guys!
Attachments
Last edited:
Toni suggested I isolate the supply reservoir caps' ground from the rest of the circuit, i.e. route these back to a segregated 'supply ground'. He also suggested changing the bypass caps to 100n from 10uF. Revised circuit and BoM attached.
I'm tempted to send this version to the board house for production while I am away for the next 10 days.
Merry Christmas everyone!
I'm tempted to send this version to the board house for production while I am away for the next 10 days.
Merry Christmas everyone!
Attachments
FWIW pcbway won't do a v-score less than 75mm. Their audit software also picks up on the fact that either side of the v-score isn't identical and labels it as two separate designs in a panel with a premium cost. I'm not going to add a v-score. I will just try to mark the centre of the PCB to provide guidance for anyone wanting to cut them in two (with Dremel, table saw or otherwise), hopefully in a manner which doesn't designate the PCB as two designs rather than one.
I'm also going to make a couple of changes. Having read a lot of Self's Small Signal Audio Design while holidaying in Florida over Christmas and New Year, I am going to change the 100nF bypass caps to be across the +/- rails rather than to ground. According to Self, the 4562 isn't fussy about decoupling and 100nF across the rails close to the package should ensure HF stability. He also makes that point that connections to ground risk coupling ground noise into the circuit. It will also save me 6 caps.
Secondly, I am going to try to fit bigger 12.5mm diameter caps for C3, C4 (and their equivalents on the other channel). A larger than 100u cap should reduce distortion below 100Hz or so.
I'm also going to make a couple of changes. Having read a lot of Self's Small Signal Audio Design while holidaying in Florida over Christmas and New Year, I am going to change the 100nF bypass caps to be across the +/- rails rather than to ground. According to Self, the 4562 isn't fussy about decoupling and 100nF across the rails close to the package should ensure HF stability. He also makes that point that connections to ground risk coupling ground noise into the circuit. It will also save me 6 caps.
Secondly, I am going to try to fit bigger 12.5mm diameter caps for C3, C4 (and their equivalents on the other channel). A larger than 100u cap should reduce distortion below 100Hz or so.
Consider the route of the output current from the opamp. Where does the +ve halfwave current eventually end up? The -ve halfwave current should also end up at the source.
Now consider if anything changes when the signal currents change very quickly. Will there be too much impedance in the return route? It's these high speed currents that use the lower impedance local decoupling.
At high frequencies the local decoupling is the source of power into the opamp and it's here that the output signal current must return.
I don't think a +ve to -ve local decoupling topology can meet this HF current route requirment.
Yep. Only the initial RF filter is connected to pin 1 chassis. All else (except the reservoir supply caps) are connected to signal ground.
I've been trying to think about current need of this little board. Quiescent (no signal) is relatively easy (I think) at 18.5-20mA per LME49740. Let's call it c60mA per channel to accommodate some leakage in the supply caps etc. But I am unsure of myself when I begin to try to estimate current need with signal added despite the fact that it is elementary topic. My presumption is that I need to make an assumption about peak line levels on the input (from my pre-amplifier) and need to know the input impedance of the IPS/VAS that this line receiver is driving. Toni, do you have the latter for your SA2014? Is it basically just the parallel resistance 47k || 8.2k or am I making a stupid mistake with all this? (Schematic excerpt attached.)
Attachments
Input impedance is mostly 47k || (8.2k+270). SA2014 likes to see a line buffer with low output impedance. Your balanced to SE circuit fulfills this need - you can expect best performance with this combination.
To drive the SA2014 to the maximum output power you need an input signal with 1.4 Vrms.
BR, Toni
Stupid question:
You're using a fancy low-noise balanced input stage only to drive a single-ended input employing a differential input stage. Why? The amplifier input might as well be modified to be balanced instead - at which point a mere buffer at the input may well get the job done. (You might be able to tell I've been reading Putzeys a lot lately.)
Polarized input coupling caps (between pull-down resistors no less)? Srsly? While this may well work for a while, you're giving the poor dielectric layer no chance to regenerate, and the poor things would eventually end up becoming rather leaky (with the specifics depending on cap quality, basic leakage and sheer luck of the draw). Better go for 22-47 µF bipolars instead, plenty big enough still. Either those or low-leakage types.
Your unbalanced input is fundamentally flawed. Or rather, having it on a device having its signal ground tied to PE is. Inputs like that rely on classic "GND think". Thou shall not have more than one ground potential. Connect to earth at one point, and keep the entire rest of the setup floating. The entire rest. Add just one device that is built as Class I or likes to develop a substantial voltage vs. earth potential (SMPS' in TVs come to mind), one safety earthed CATV or antenna connection, one shielded network cable that's carrying earth potential from somewhere, and you're screwed - there's your ground loop. In that sense, consumer and pro audio are fundamentally incompatible (and consumer audio is just broken these days).
Putzeys suggests one way of implementing an unbalanced input in his "Dealing with legacy pin-1 problems" article. I would add a ground lift switch to that, just in case the source's common mode potential likes to drift way off (at that point a transformer is the best option anyway). You may also decide not to add the unbalanced input at all and make proper adapter cables instead, complete with impedance balancing (Bill Whitlock has written a lot on the subject, Putzeys also describes a version though the RC for the cable shield may be better off being connected on the transmitting end).
BTW, having a super-low-noise input is all good and well, but have you looked at the output noise of your preamp, and how far down you are likely to ever get it? If the balanced output likes to double the level (as it often does), have you considered going for an input gain of -6 dB?
Same story as the O2 headphone amp's input stage. I'd be pretty hard-pressed to name any source that got anywhere near its noise levels. At the same time, the poor little NJM2068 may break a bit of a sweat when confronted with high input levels due to feedback network impedance, depending on gain setting.
You're using a fancy low-noise balanced input stage only to drive a single-ended input employing a differential input stage. Why? The amplifier input might as well be modified to be balanced instead - at which point a mere buffer at the input may well get the job done. (You might be able to tell I've been reading Putzeys a lot lately.
)Polarized input coupling caps (between pull-down resistors no less)? Srsly? While this may well work for a while, you're giving the poor dielectric layer no chance to regenerate, and the poor things would eventually end up becoming rather leaky (with the specifics depending on cap quality, basic leakage and sheer luck of the draw). Better go for 22-47 µF bipolars instead, plenty big enough still. Either those or low-leakage types.
Your unbalanced input is fundamentally flawed. Or rather, having it on a device having its signal ground tied to PE is. Inputs like that rely on classic "GND think". Thou shall not have more than one ground potential. Connect to earth at one point, and keep the entire rest of the setup floating. The entire rest. Add just one device that is built as Class I or likes to develop a substantial voltage vs. earth potential (SMPS' in TVs come to mind), one safety earthed CATV or antenna connection, one shielded network cable that's carrying earth potential from somewhere, and you're screwed - there's your ground loop. In that sense, consumer and pro audio are fundamentally incompatible (and consumer audio is just broken these days).
Putzeys suggests one way of implementing an unbalanced input in his "Dealing with legacy pin-1 problems" article. I would add a ground lift switch to that, just in case the source's common mode potential likes to drift way off (at that point a transformer is the best option anyway). You may also decide not to add the unbalanced input at all and make proper adapter cables instead, complete with impedance balancing (Bill Whitlock has written a lot on the subject, Putzeys also describes a version though the RC for the cable shield may be better off being connected on the transmitting end).
BTW, having a super-low-noise input is all good and well, but have you looked at the output noise of your preamp, and how far down you are likely to ever get it? If the balanced output likes to double the level (as it often does), have you considered going for an input gain of -6 dB?
Same story as the O2 headphone amp's input stage. I'd be pretty hard-pressed to name any source that got anywhere near its noise levels. At the same time, the poor little NJM2068 may break a bit of a sweat when confronted with high input levels due to feedback network impedance, depending on gain setting.
Last edited:
Lots of points here. Let me respond quickly to a few of them now and perhaps more in the morning. It's late and I didn't get any sleep last night courtesy of a red-eye transatlantic flight with a very young baby (one that is now struggling to adjust to a return to UK timezone).
I don't get to modify the amplifier - I'm using Toni's existing and superb "SA2014" design with his boards. Having read Self, whose design this is (well, the balanced side at least to which I have tried to integrate his guidance re unbalanced line inputs) it seems "a mere buffer" would have a noise level substantially worse than an unbalanced connection and it was this which provided the motivation for his design - to lower the noise of the balanced input buffer to at least equal that of an unbalanced connection.
The input coupling caps are not polarised. I've just used a stock Eagle part in the schematic and board. There will be no + marking on the board. The caps to be used are UEP1V471MHD bipolar 35V 470u. (The BoM I posted listed the smaller 100uF ones, but the decision was made to provide for the larger caps.)
Signal GND is not tied to PE. The only connection to PE via XLR pin 1 is the initial RF filter, the idea being to not pollute signal ground with any RF noise. (The GND of the supply caps returns to the supply. All other connections are to an isolated Signal Ground.)
As you will have noted from earlier posts, I am personally less concerned with single-ended input but added it anyway for 'flexibility'. Hence, I am not concerned about the volume difference that will exist from the two inputs. I don't expect (nor expect anyone else that might decide to use it) to bounce between inputs and so find the relative volume levels annoying. So I did not worry about a typology where balanced signals had a gain of -6dB while unbalanced signals had unity gain and stuck with unity gain for both. It made integrating the unbalanced signal much easier. Of course it could be changed so that the unbalanced connection skirts the low noise unity gain balanced input buffer (which in turn could be altered for -6dB gain) but, as said, I wasn't fussed about the relative volume issue. The unbalanced signal with RF filtering components etc could then connect directly at the output connector or have its own fancy low noise input buffer. When not concerned about relative gain it made sense to use the typology shown here. Perhaps there is a typology that readily integrates Self's low noise balanced input buffer (see Fig 18.28 is Small Signal Audio Design 2nd ed) with different gain structures for each signal type (see Fig 18.10) but it didn't seem to me to be worth pursuing.
My preamplifier specs list THD +N @ 1kHz as less than 0.0006% at all output levels. But, regardless, the idea was to build an exceedingly good balanced input for use with Toni's amp.
I don't get to modify the amplifier - I'm using Toni's existing and superb "SA2014" design with his boards. Having read Self, whose design this is (well, the balanced side at least to which I have tried to integrate his guidance re unbalanced line inputs) it seems "a mere buffer" would have a noise level substantially worse than an unbalanced connection and it was this which provided the motivation for his design - to lower the noise of the balanced input buffer to at least equal that of an unbalanced connection.
The input coupling caps are not polarised. I've just used a stock Eagle part in the schematic and board. There will be no + marking on the board. The caps to be used are UEP1V471MHD bipolar 35V 470u. (The BoM I posted listed the smaller 100uF ones, but the decision was made to provide for the larger caps.)
Signal GND is not tied to PE. The only connection to PE via XLR pin 1 is the initial RF filter, the idea being to not pollute signal ground with any RF noise. (The GND of the supply caps returns to the supply. All other connections are to an isolated Signal Ground.)
As you will have noted from earlier posts, I am personally less concerned with single-ended input but added it anyway for 'flexibility'. Hence, I am not concerned about the volume difference that will exist from the two inputs. I don't expect (nor expect anyone else that might decide to use it) to bounce between inputs and so find the relative volume levels annoying. So I did not worry about a typology where balanced signals had a gain of -6dB while unbalanced signals had unity gain and stuck with unity gain for both. It made integrating the unbalanced signal much easier. Of course it could be changed so that the unbalanced connection skirts the low noise unity gain balanced input buffer (which in turn could be altered for -6dB gain) but, as said, I wasn't fussed about the relative volume issue. The unbalanced signal with RF filtering components etc could then connect directly at the output connector or have its own fancy low noise input buffer. When not concerned about relative gain it made sense to use the typology shown here. Perhaps there is a typology that readily integrates Self's low noise balanced input buffer (see Fig 18.28 is Small Signal Audio Design 2nd ed) with different gain structures for each signal type (see Fig 18.10) but it didn't seem to me to be worth pursuing.
My preamplifier specs list THD +N @ 1kHz as less than 0.0006% at all output levels. But, regardless, the idea was to build an exceedingly good balanced input for use with Toni's amp.
Last edited:
THAT is 'dated'? Do you know how long 5532s have been around? I was using them almost 35 years ago. They're fine chips and LOTS of audio has been run through them but given the choice, the THAT 1240, Analog Devices SSM2141 or Burr-Brown INA134 with their laser trimmed resistors are hard to beat and will almost certainly be better balanced than a discrete solution. At line level you'd be very hard pressed to find any significant noise with any of those chips.
Not to be nasty but I have a bad time reading pictorial 'schematics' and it's getting worse as I get older. I much prefer a traditional schematic.
G²
Not to be nasty but I have a bad time reading pictorial 'schematics' and it's getting worse as I get older. I much prefer a traditional schematic.
G²