THAT 1646 oscillation problem (solved !! )

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

Freq Band

Well-known member
GDIY Supporter
Joined
Jan 5, 2006
Messages
608
Location
Electra City
I just built my first THAT 1646 balanced driver circuit. (with the 10uf NP caps between sense and out(s)). Point-to-point wired:

DSCN2556.jpg




I don't understand the results (oscillation?)......

DSCN2553.jpg


(bottom) 1khz sine direct from function gen.
(upper two) 1646 outputs.

Do these not like to be socketed?
info:
+V bypassed with ceramic 100nf's
NP 7.2 uf's measure out at 10.2 uf
Grounded everywhere.

Am I forgetting caps/resistors on the inputs or outputs?
The ground pin on the 1646....is this signal ground only, power ground, or either, or both?

=FB=
 
Yes, the bypass caps are connected to the power supply ground, but the 1646's pin3 ground was paired with the signal ground, so there is about 1 foot of that wire until star ground. .....in addition to the 1 foot of the board's "power input" ground wire to star.
When I get home, I'll tie all grounds together locally, on the board.

I'll also test it with a 50k pot on the input in a standard "volume" configuration...but only for a test.

thanks!!

=FB=
 
SUCCESS !!

Now, all grounds on board tied to one point.

dscn2561.jpg


When installing this circuit, output XLR pin 1 would usually be connected to..
All other XLR's pin 1 ?
The ground point on this driver board ?
or..?


If you say "depends".....then consider this:

My PSU uses a floating ground,
Case is tied to safety earth,
I want to mute each input (including it's ground) when not in use, via relays.

(sorry, ground is usually a hurdle for me :? )
BTW, this is for a HiFi preamp, that might occasionally see studio gear attached, but for the most part, it will be for monitoring my mixes, and pleasure listening.


=FB=
 
XLR pin 1 would usually be connected to..

Ah, "ground", JR, truer words are rarely spoken.

There are more than a couple of flavors of ground, and it can definitely help to keep track of who's who and when and where they want to play together, or not. The XLR pin 1, as well as the TRS sleeve, connection is not signal ground. It therefore should not be lumped in with the PS and signal grounds inside the gear. Pin 1 is the shield drain. If you attach it to your signal ground on the PC board in your gear, any current that the shield is trying to shunt will transfer to said ground. Shield should connect to shield, and the chassis is the gear's shied. Pin 1 should connect to the chassis directly. Do not pass Go, do not collect $200. In fact, there are many pieces of professional gear who's noise performance can be markedly improved by severing the PC mount XLR pin 1 connection to the ground plane and wiring them straight to the chassis. Don't ask how I know.

Neil Muncy's paper (preprint 3930) presented at the 97th AES convention (AES Journal 6/95) "Noise Susceptibility in Analog + Digital Signal Processing Systems" is one of the definitive bits on this. In it, he defines the "Pin 1 problem". This is what one gets by attaching pin 1 to the ground plane, or signal ground, like the aforementioned manufacturers. Via common impedance coupling, all sorts of nasties can show up in the audio. Kind of like daisy chaining power strips together or shorting the neutral and ground in an iso ground AC run.

I read a more recent interview with him where he mentions that the manufacturers have gotten hip to it, and the problem is becoming less prevalent. I just have to say, he must only work with nice toys. So many mid level and "prosumer" bits of gear are done so far on the cheap, they wouldn't connect anything at all if they didn't have to, much less if doing it right cost 3 cents.

Anyway, that paper, and Kenneth Fause's "Fundamentals of Grounding, Shielding and Interconnection" from the same convention will school one proper on this sort of thing. I can't find them online, though I'm sure you could buy them from AES should the fancy take you. Quite a few good treatises footnote one or the other, though. A search on "pin 1 problem" should provide a good bit of reading.

I want to mute each input (including it's ground)

Balanced signals don't really have a ground, per se. (Somebody's going to kill me for that, I'm sure). The signal is the difference between the "hot" and "cold" legs, rather than the difference between "hot" and "ground", "cold" and "ground", or some combination, so I wouldn't worry too much about switching the shield out of circuit unless you're having other interference troubles.

Anyway, long story short, (I know, too late.) pin 1 to chassis, and it's not out of the question to lift one side in the cable. The only XLRs that absolutely need shield connected through are ones used to phantom power microphones.
 
Very interesting...
But if you don't connect circuit ground to chassis and if a circuit fault happens, and a power supply rail connects to chassis ground, does the IEC fuse blow ?

[edit]
I just realized that it is only the XLRs pin 1 that are not connected to the circuit ground directly following your theory, the circuit ground is still connected to the chassis at one point, right ?
 
keefaz,

Yeah, it's about the shield. Sorry about the rant, but it's a pet peeve.

All the grounds eventually connect to the same place, as do the neutral and U-ground in an iso ground AC system, at the last possible point. If they connect upstream, any common impedance conspires to shovel noise from one into the other. I know places that branch their U-ground wires in different gauges to deal with this. I prefer as many home runs to the box as I can get, and then to go huge and short from box to stake, but hey.

From what I gather, the problem became widespread when gear started using PCB mount connectors. When they were all chassis mount, it was easier for the manufacturers to do it right anyway. So they did. With PCB mount, it's easier just to have the ground plane meet pin 1, and not bother running a wire to the chassis. I've chased crazy noises down to exactly this problem in even what is considered pretty nice gear. The fix is easy, but really, should we have to?
 
Without getting into a full tutorial on grounding, IMO this (pin 1) became a problem when audio transformer i/o fell out of popular use due to cost. Transformers are very forgiving of sloppy interface practices.

Over the last 10-15 years awareness of the "pin 1" problem has helped many in error to see the correct path, and modern gear is generally better behaved.

Ground is most confusing if you try to think of it as anything other than another signal node getting connected to and passed around.

For safety bonding that ground path need to pass enough current to trip circuit breakers under fault conditions. To prevent noise and fault currents from corrupting audio signals, signal grounds becomes local reference nodes that may touch that system ground at one point but do not actually carry ground current. Differential audio signals are commonly referenced to or from these local signal grounds.

To bring this full circle back to the original thread topic, for stability some circuits require a low impedance HF bond between local signal ground and power or system grounds.

JR
 
[quote author="JohnRoberts"]
Ground is most confusing if you try to think of it as anything other than another signal node getting connected to and passed around.

For safety bonding that ground path need to pass enough current to trip circuit breakers under fault conditions. To prevent noise and fault currents from corrupting audio signals, signal grounds becomes local reference nodes that may touch that system ground at one point but do not actually carry ground current. Differential audio signals are commonly referenced to or from these local signal grounds.
JR[/quote]
This should be in the meta IMO
Very well explained, thank you!
 

Latest posts

Back
Top