Signal isolation, transformer coupling and optos

[defaced]

I'm working on a project where I'd like to have an isolated audio output. This is solely to break ground loops in connected mains powered equipment. I've got a preamp and a power amp for instance, I'd like to isolate the output of the preamp to break any signal/mains ground loops. I see a couple of ways to do this and have a couple of questions about some of the ways:

• Transformer coupling. Looks like you can get 200Hz to 15kHz low powered audio transformers inexpensively enough, but the 200 to 15k bandwidth is kind of a bummer, I'd prefer something flatter, but I'm willing to experiment since these types of transformers are used in other guitar circuits.
• Capacitor coupling. If you couple the signal and ground of a stage do you break ground loops? I'm thinking "no" because while you are AC coupling the signal you are still AC coupling ground, so it's a moot point
• Optoisolators. In the industrial equipment I use (advanced mechanized welding power supplies), the I/O is isolated using optoisolators to keep welding round (hundreds of amps) away from signal and small motor grounds (2 amps tops). However, it seems because of ROHS, most optos are some composition of solid state devices inside and I'm not sure that would play with audio very well (thinking along the same lines as solid state relays, digital pots and the like).
• Something I'm not thinking about behind door number 4...?

Now since I'm going to have a buffered output, something tells me that I have to pay attention to the details when selecting an isolation device to preserve the low impedance nature of the buffer, yes?

As always, thanks for any thoughts or considerations. Seems the deeper I go down the hole the more I see that needs to be considered.

[guitician]

Here is the industry standard for audio transformers...Lundahl Transformers - General purpose transformers

They are not cheap though.... Do you need isolation? Or are you just trying to avoid ground loops? Coupling a preamp to a power amp using a transformer is never done as far as I know. Except back in the early days of tube amplifiers.
Optoisolators wont pass linear audio as far as I know....

[rjb]

• Optoisolators.... However, it seems because of ROHS, most optos are some composition of solid state devices inside and I'm not sure that would play with audio very well...

I don't understand the question/concern.
AFAIK, optoisolators have always contained solid state devices- an LED and a transistor.
I don't know what ROHS has to do with it.
Are you thinking optos are like CMOS switches?
Do you want to avoid "non-historically correct" SS devices in a tube amplifier?

Later,
Ralph "Not Getting It" Barthine

[defaced]

Thanks for the link guitician, I'll check those out.

AFAIK, optoisolators have always contained solid state devices- an LED and a transistor.
I don't know what ROHS has to do with it.

As I understand the history (which is likely wrong/only partially correct, this is just my hobby and I'm relatively new at it), optoisolators were at one point in time made with a LED and a LDR (for instance the Vactrol brand optos that are often used for channel switching in amps). Because of the materials required to make an LDR, these have been phased out for types of devices (wiki lists several in addition to LRDs). I could care less about the presence of taboo SS devices in my amp provided they do their job adequately. My concern is that I know squat about the performance of the different types optos or if they'd even be appropriate for audio circuits, and if so which kind and what parameters to look for, thus the question.

[rjb]

...optoisolators were at one point in time made with a LED and a LDR...

Maybe I'm confusing "optoisolators" with "optocouplers".
I've only used optocouplers for 20mA comm loops, not for audio.
But I do know they've been used in the tremolo ckts of some SS amps.

Here are some excerpts from a text book:
An optocoupler uses an LED optically coupled to a photodiode or a phototransistor in a single package.
A key parameter...is the CTR (current tranfer ratio).
The CTR is an indication of how efficiently a signal is coupled from input to output...
expressed as the ratio of a change in the LED current to the corresponding change in the... phototransistor current.
It is usually expressed as a percentage.
Figure 4-33....CTR vs forward LED current....
The figure shows that the typical optocoupler could exhibit pretty non-linear response, depending on how it is biased.

As guitician queried, are you sure you need to isolate your preamp from power amp?

-rb

[Steve Conner]

There's a schematic for an isolated buffer out there, made with transformers. I can't remember who designed it. Could have been RG, or Jensen. But it is tested in the guitar application.

To me at least, optoisolator and optocoupler are the same thing. All optoisolators will pass audio to some extent, except the LDR kind which are way too slow. There is a specific kind from HP/Agilent/Avago designed to pass analog signals with low distortion. But they all have the fundamental problem that the receive side photocurrent is tiny and needs amplified to drive an audio line. To preserve isolation, the amplifier then needs an isolated power supply, which increases complexity. I've built galvanically isolated RS232 ports this way.

Last but not least, the audio balanced line driver and receiver chips from TI and That. The driver emulates a floating transformer winding using analog voodoo. Either the driver or receiver will break a ground loop by itself. If you use both, with a 3-pin XLR cable, you can get excellent results.

I've used them before, but I forgot the part numbers.

[defaced]

As guitician queried, are you sure you need to isolate your preamp from power amp?

We'll find out. I was rooting around in the preamp I'm building last night and noticed I didn't have the output grounded the way I had intended (this is what happens when you take a 3 month break from a project), so this is somewhat preemptive, but if the CYA is cheap/easy/etc, the extra part or 5 isn't going to kill me.

There's a schematic for an isolated buffer out there, made with transformers. I can't remember who designed it. Could have been RG, or Jensen. But it is tested in the guitar application.

Yep, those are the 200 to 15k bandwidth transformers I mentioned in my first post. Actually, the ones in RGs schem are lower bandwidth, maybe 300 to 3.5k (I've even recommended them as a possibility for a project as they are used in the BYOC signal splitter). Either way, I'm definitely into looking at them, just wanted to see if there are other options out there that were better in terms of bandwidth while maintaining the reasonable cost.

I'll look into those TI and THAT chips. Thanks Steve.

[loudthud]

Those transformers that specify 200Hz to 15KHz are specified at maximum power, usually 100s of milliwatts. At one milliwatt the bandwidth is much greater. RG usues two in his signal splitter A/B box that you will find on his website.

[defaced]

At one milliwatt the bandwidth is much greater

Bang on, that's exactly the nugget of info I wasn't seeing. Knew I had to be missing something simple. Thanks LT!

[R.G.]

Beyond the lower-power/higher-bandwidth thing, you can also extend the low end by driving the primary side from a lower impedance than "matched". The low end limitation is generally the primary inductance's impedance "shorting out" the reflected secondary load. If you drive the primary from a low impedance that can supply a lot of current into the primary inductance, the primary inductor can eat more current, but the reflected secondary load still gets low-loss signal voltage, so the total bandwidth through the transformer is extended.

It's a trick to get extended low frequency response out of cheaper transformers, and seems to work. The $3.00 300-3000Hz transformers give 60Hz to 22kHz on test with a low impedance driver.

[loudthud]

To find the project I referenced, go to: New Page 1

On the left click FX Projects

Then in the second column click Hum Free Signal Splitter

The transformer he uses is Mouser 42TM018

[kg]

i sidestepped this issue by running balanced outputs from preamp to poweramp, and grounding only the preamp side of the conductor shield.

you get that kind of flexibility when you roll your own...

[Gregg]

OEP and Triad transformers have better specs and are not much more expensive:

Audio Transformer | Audio Frequency Transformers | Farnell United Kingdom | Results Page 2

[dai h.]

is there really a need for the transformer? It would seem to me if the gear has correct grounding schemes (should be doable if DIY'ed) an unbal. interconnection won't cause a hum prob., though I suppose it would help if the goal is to ensure that the ground can be broken safely (for gear that can't be modified (pricey 'vintage correct' gear that you don't own, etc.) or modded easily (no time)). From what I understand RG's example(GEO transformer splitter project) is the cheapest way to do it (there is a commercial one from Suhr Tech. IIRC that is similar), that is use inexpensive ones wound for lower impedances (the wire isn't as thin as for higher Z ones w/good freq. response which need to be more elaborately designed/constructed to minimize capacitance between windings = more expensive), interface them with buffer to get high Z input, drive them with low Z to extend low freq. response (I think of the pri. as an inductor/coil winding sort of like a lower Z tube input). Basically (AIUI) the impedance rating of the transformer means that it's optimized to work at those Zs (and specified levels) and generally will work pretty well a bit outside of that impedance range but you don't want to go too far off (so 600/600 might be ok as a 300/300 or 800/800 but don't try to use it as 1Meg/1Meg). (Plus the aforementioned about driving the Pri. w/lower Z which could help but may not always.) Also, putting transformers in a signal path can make the result a bit of an unknown unless there is some understanding. (Again, from my limited understanding) a Xfrmr. is like two coil windings side by side with each having a resonance (L+C elements make a resonant "tank circuit") which can cause "ringing"(high freq. resonance) hence the need for a correctly terminated secondary (using an R, a Zobel, or both--assuming flat response is the goal--to damp the resonant peaks to flatten towards a more ideal response). Otherwise, phenomenon such as thin weird sound, distortion of the next stage (from the high amplitude ringing intermodulating w/lower freqs.) and different seemingly unpredictable results w/different input Zs can occur.

[tubegeek]

The way I look at it, defaced is looking to add another transformer to his rig. More transformers is ALWAYS better!