New PCB for shoebox size NAC preamp part II
- Thread starter timH
- Start date
- Status
chiily
PFM Special Builder
Fatmarley
"It appears my intelligence circuits have melted"
So I've built a starfish. I've got a traco on board, and it's fully symmetric. I've got the chokes and everything fully populated. Last night I had sine and square waves going in and out of it.
Today, no joy. The voltage output of the traco is +/-6.5V instead of +/-15V. If
I unhook either one of the chokes filtering the + and - rails, the traco output goes back to +/-15V. Put them back in circuit and it's back to +/-6.5V.
I tried removing the NERT and using a piece of wire from traco 0V to power 0V. Still +/-6.5V. All of the 14 on-board regs are giving an output of approx +/-5.25V each, as expected.
What gives? Any ideas?
Today, no joy. The voltage output of the traco is +/-6.5V instead of +/-15V. If
I unhook either one of the chokes filtering the + and - rails, the traco output goes back to +/-15V. Put them back in circuit and it's back to +/-6.5V.
I tried removing the NERT and using a piece of wire from traco 0V to power 0V. Still +/-6.5V. All of the 14 on-board regs are giving an output of approx +/-5.25V each, as expected.
What gives? Any ideas?
martin clark
pinko bodger
Something odd going on with the effective LC filter vs. what the SMPS modules internal control loop expect? Do the rails still collapse if you parallel the inductors with 100, 10, 1R ?
Not sure. I'll check though.
After further experimentation, I've found that everything behaves perfectly with only the + rail connected by unlinking the differential choke. If I swap things over, leaving only the - rail running, then the SMPS output is +15V, -9V. The 9V fluctuates by a margin of around 20-30mV, continually rising and falling.
I assume the smps is happy with low-ESR caps hanging off the output? I've got 470pF across +/0V and -/0V right on the output pins. Then there's a pair of 56uF/25V oscons across +/0V and -/0V, with a wima 10u MKS2 film across +/-. After that there's a common mode choke (two windings on a toroid, one winding per rail), followed by another oscon/oscon/wima triplet. Then there's the diff chokes (two toroids, one winding each, one per rail) followed by another pair of oscons and a 220u/50V panasonic fm across +/-. After that, all the local regs come into play.
After further experimentation, I've found that everything behaves perfectly with only the + rail connected by unlinking the differential choke. If I swap things over, leaving only the - rail running, then the SMPS output is +15V, -9V. The 9V fluctuates by a margin of around 20-30mV, continually rising and falling.
I assume the smps is happy with low-ESR caps hanging off the output? I've got 470pF across +/0V and -/0V right on the output pins. Then there's a pair of 56uF/25V oscons across +/0V and -/0V, with a wima 10u MKS2 film across +/-. After that there's a common mode choke (two windings on a toroid, one winding per rail), followed by another oscon/oscon/wima triplet. Then there's the diff chokes (two toroids, one winding each, one per rail) followed by another pair of oscons and a 220u/50V panasonic fm across +/-. After that, all the local regs come into play.
Super puzzled now!
I decided to isolate individual -ve regulators and see if the problem went away: there are only six on the board so it's not hard to do. I chopped the input rail to one of the regs, then measured the traco output... -15V
...but. I didn't want to be too sure. So I chopped another -ve rail to another -ve reg, re-connected the previous regulator, and powered up... -15V!!! What the f**k?
It appears that no matter which of -ve regs I chop out, as long as I chop out one, the -ve rail comes back up to voltage....
...more investigation to be done, that's for sure.
I decided to isolate individual -ve regulators and see if the problem went away: there are only six on the board so it's not hard to do. I chopped the input rail to one of the regs, then measured the traco output... -15V
...but. I didn't want to be too sure. So I chopped another -ve rail to another -ve reg, re-connected the previous regulator, and powered up... -15V!!! What the f**k?
It appears that no matter which of -ve regs I chop out, as long as I chop out one, the -ve rail comes back up to voltage....
...more investigation to be done, that's for sure.
Carl,
There's a component error or a short somewhere on your board.
Check, double check and triple check all the component values and go over your board with a magnifying glass.
Try also disconnecting all the 47R output series resistors on the local regs.
This disconnects the local regs from the audio circuits - if the regs then all work fine the problem is in the audio circuit somewhere.
Cheers,
Jim.
There's a component error or a short somewhere on your board.
Check, double check and triple check all the component values and go over your board with a magnifying glass.
Try also disconnecting all the 47R output series resistors on the local regs.
This disconnects the local regs from the audio circuits - if the regs then all work fine the problem is in the audio circuit somewhere.
Cheers,
Jim.
Yup, definitely sounds like a solder short or a component error somewhere on your board.
The traco will auto-shutdown to prevent damage when presented with a load it cannot drive.
A functioning starfish is well within the current supply capabilities of a traco.
It's quite nice for debugging as well - the traco safety shutdown prevents damage - if you had been powering this from something like a hicap, smoke might have been released
Let us know how you get on.
You have my email address if you need more help.
Cheers,
Jim.
Thanks Jim. I'll hook the 337s back into the circuit, then lift a leg on all of the 27Rs (I'm not actually using 27Rs, but rather 100uH chokes with 4ohm series resistance, but that shouldn't make any difference).
I'll hook it up and see what happens. If all is well, I'll start adding them all back one at a time until something breaks.
I wish I had time to do it today, but Mrs Hacker has plans for yard work that trump plans for Starfishes!
I'll hook it up and see what happens. If all is well, I'll start adding them all back one at a time until something breaks.
I wish I had time to do it today, but Mrs Hacker has plans for yard work that trump plans for Starfishes!
I had a quick look at the board over a cup of coffee. I discovered a bit of a brain fart: you know the 14 47uf decoupling Stargets (ok, I'm using Panasonic FMs, but still)... anyway, one of the gain stage -ve rail FMs has its polarity reversed. I suspect that cap has given up and is making the circuit behave in weird and wonderful ways... I'll whip it out and replace it
Well bugger me if I haven't fixed it...
...and the problem wasn't the Starfish in the first place! It turns out the upstream PSU was incapable of delivering the necessary 250mA needed by the Traco! I was using one rail of a Twisted Pear LCDPS, and I switched to an old laptop supply on a hunch. Bingo! Everything worked as normal. Switching back to the LCDPS... nope, not enough power. So I configured both rails of the LCDPS to exactly 12.00V and hooked them up in parallel to the Traco... and it worked!
I've spent hours on this problem, swapped many caps (that's eBay tants for you) and scratched my head in bewilderment, all the while overlooking the obvious. At least it's fixed now
...and the problem wasn't the Starfish in the first place! It turns out the upstream PSU was incapable of delivering the necessary 250mA needed by the Traco! I was using one rail of a Twisted Pear LCDPS, and I switched to an old laptop supply on a hunch. Bingo! Everything worked as normal. Switching back to the LCDPS... nope, not enough power. So I configured both rails of the LCDPS to exactly 12.00V and hooked them up in parallel to the Traco... and it worked!
I've spent hours on this problem, swapped many caps (that's eBay tants for you) and scratched my head in bewilderment, all the while overlooking the obvious. At least it's fixed now
I wired in the attenuator and ran some sine waves through it, and everything checked out ok on the 'scope. It's now hooked up to my rig and wow does it sing! I was just going to test it, but I think I'll leave it hooked up and running in for a while. Here's how she looks just now:
You'll see a few non-BOM things in that pic:
- Heaps of 10uf/63V axial tants that are doing duty as regulator input/output/adjust bypasses.
- Heaps of Wima MKS2-XL that are doing duty as current source decoupling, input and output coupling, and feedback. You'll also see a couple in the Traco supply chain, although most of the caps have been pulled from there at the moment. They'll go back in soon.
- There are no regulator output 27R or inductors. I'm using wire links just now.
- Panasonic FM 47u/25V doing audio circuit decoupling (instead of 47u Elna Stargets). They're ultra-low ESR, high-ripple audio grade caps. Apparently. Normally ultra-low ESR caps on the output of LT1086 and LM337 regs would be a recipe for disaster, but there's a good bit of series R provided on the reg outputs by the axial tants. That keeps things nice and stable while providing a nice low impedance decoupled power source for the audio circuits.
- Vishay/Dale Rn55 and RN60 resistors throughout except for:
- Welwyn rc55y feedback resistors (and the occasional other 100k and 10k where I had available spares)
The CLCLC filter will look like this when finished:
This version of the Starfish is dynamic! It's got wonderfully detailed highs and very snappy mids, and super-tight bass. I'm going to leave it running in for a while before making my mind up, but my initial impression is highly favourable
You'll see a few non-BOM things in that pic:
- Heaps of 10uf/63V axial tants that are doing duty as regulator input/output/adjust bypasses.
- Heaps of Wima MKS2-XL that are doing duty as current source decoupling, input and output coupling, and feedback. You'll also see a couple in the Traco supply chain, although most of the caps have been pulled from there at the moment. They'll go back in soon.
- There are no regulator output 27R or inductors. I'm using wire links just now.
- Panasonic FM 47u/25V doing audio circuit decoupling (instead of 47u Elna Stargets). They're ultra-low ESR, high-ripple audio grade caps. Apparently. Normally ultra-low ESR caps on the output of LT1086 and LM337 regs would be a recipe for disaster, but there's a good bit of series R provided on the reg outputs by the axial tants. That keeps things nice and stable while providing a nice low impedance decoupled power source for the audio circuits.
- Vishay/Dale Rn55 and RN60 resistors throughout except for:
- Welwyn rc55y feedback resistors (and the occasional other 100k and 10k where I had available spares)
The CLCLC filter will look like this when finished:
This version of the Starfish is dynamic! It's got wonderfully detailed highs and very snappy mids, and super-tight bass. I'm going to leave it running in for a while before making my mind up, but my initial impression is highly favourable
Hi Carl,
Glad you got it working - looking good!
I'm a bit worried about:
You dont have any series resistance between the local regulator and the FM cap - other than the resistance of your wire links.
The FM cap is currently in parallel with the tant.
What you describe as 'dynamic' could be mid-band resonance on the output of the local regulators - just causing a peaking in those mid-range frequencies that make things sound fast, alive, and ..well.. dynamic. You dont work for Naim do you?
I reccommend getting a plane-jane 47R resistor in there in place of the wire links - the local regulator will only see the tant and therefore be happy, the audio circuit will see the FM cap and may also be happy. Report back how it sounds
Once again - looking good. More pron please.
Jim.
Glad you got it working - looking good!
I'm a bit worried about:
You dont have any series resistance between the local regulator and the FM cap - other than the resistance of your wire links.
The FM cap is currently in parallel with the tant.
What you describe as 'dynamic' could be mid-band resonance on the output of the local regulators - just causing a peaking in those mid-range frequencies that make things sound fast, alive, and ..well.. dynamic. You dont work for Naim do you?
I reccommend getting a plane-jane 47R resistor in there in place of the wire links - the local regulator will only see the tant and therefore be happy, the audio circuit will see the FM cap and may also be happy. Report back how it sounds
Once again - looking good. More pron please.
Jim.
Hi Jim,
I've had my SB3 on random for several hours now, giving the 'fish time to bed in a bit and give me a feel for all genres of music. It sounds bloody marvelous. Thank you!
With regard to the series R, I would have thought that mid-band peaking would have been immediately obvious in things like string instruments and especially female vocals, yet I hear no such thing (yet). That Naim "quality" of exaggerated (ok, peaky and harsh) mids and upper mids is noticable by its absence. Quite the opposite in fact. Björk is currently playing quite loud and her voice is like honey, as was Natalie Merchant's, Joni Mitchell's, Erykah Badu's...
The amp sounds so good that I'm actually reluctant to fiddle with it - and that must be a first! However, at some point I will try and add a little series R - I've got some 33R kicking around that would do the job. I also need to get the CLCLC filter back in place.
One thing I notice is that the NERT gets bloody hot on the rail-facing PNP transistor (I used 2x D44H11 and 1x D45H11 with 910R, 100R, 200R, and 13R as per NERT spec). So hot in fact that I can't touch my finger to it for more than a second or so. The NPNs are ok, although the positive rail one is pretty warm. The PNP is frikkin roastin'!
I downloaded a siggen for the PC and wanted to try and see starfish's effect on the leading edges of square waves, but my computer soundcard is so shit that there's a whole crap-load of under/over shoot on the output wave form and there's no way I'd get an accurate picture of the starfish from it. Shame, because that would have been a great way to optimize the cct - stick in a 100R trimmer in place of the 27R and trim away until perfect.
Cheers!
I've had my SB3 on random for several hours now, giving the 'fish time to bed in a bit and give me a feel for all genres of music. It sounds bloody marvelous. Thank you!
With regard to the series R, I would have thought that mid-band peaking would have been immediately obvious in things like string instruments and especially female vocals, yet I hear no such thing (yet). That Naim "quality" of exaggerated (ok, peaky and harsh) mids and upper mids is noticable by its absence. Quite the opposite in fact. Björk is currently playing quite loud and her voice is like honey, as was Natalie Merchant's, Joni Mitchell's, Erykah Badu's...
The amp sounds so good that I'm actually reluctant to fiddle with it - and that must be a first! However, at some point I will try and add a little series R - I've got some 33R kicking around that would do the job. I also need to get the CLCLC filter back in place.
One thing I notice is that the NERT gets bloody hot on the rail-facing PNP transistor (I used 2x D44H11 and 1x D45H11 with 910R, 100R, 200R, and 13R as per NERT spec). So hot in fact that I can't touch my finger to it for more than a second or so. The NPNs are ok, although the positive rail one is pretty warm. The PNP is frikkin roastin'!
I downloaded a siggen for the PC and wanted to try and see starfish's effect on the leading edges of square waves, but my computer soundcard is so shit that there's a whole crap-load of under/over shoot on the output wave form and there's no way I'd get an accurate picture of the starfish from it. Shame, because that would have been a great way to optimize the cct - stick in a 100R trimmer in place of the 27R and trim away until perfect.
Cheers!
sq225917
Bit of this, bit of that
If Bjork and Joni sound like Honey then you have big issues.... lol.
- Status