« Load-Current Margins | Main | Minimum Power Supply Load »

September 05, 2000

Isolation Transformer

Question: Can you explain an isolation transformer?

Original Question: Can you kindly tell me briefly the theory of the "isolation transformer" and how does it reduce coupled noise and how does it provide better isolation rather than other transformers between primary and secondary windings and does it protect the circuit from transient over voltage, sags, harmonics, and radio frequency interference in power supplies.

Answer: The following assumes a line frequency transformer. All transformers provide dc isolation between isolated windings. However, because they all have capacitance from a winding to any nearby conductor, such as other windings, the core, and the chassis, they can couple high frequency signals and noise to these nearby conductors. This capacitive coupling mechanism can carry high frequency noise on the input primary to the output secondary windings or vice versa.

Isolation transformers are designed to minimize this coupling by construction that minimizes capacitive coupling, or by the use of shields, or both.

Shields usually increases the capacitance between primary and secondary if the shield is not connected. However, by connecting the shield to the appropriate place, the noise can be diverted to a harmless place. Theoretically, you need one shield per system ground, although you can often get by with one less. For example. If you had a primary ground, a secondary ground, and a safety (chassis) ground, you would need three shields. One is connected back to primary ground, one to secondary ground and one to chassis ground. Since you may not have to protect secondary noise from the primary, or vice versa, you might be able to get by with two shields. Many transformers only have one shield and it is difficult to connect this so it has the desired affect. If both the primary and secondary windings are in a complete shield, the transformer is called "box" shielded and this is one of the highest quality and more expensive shielding method.

It is very important how the shields are connected to their respective grounds. Any connecting wire contains inductance and this sets up a resonant circuit that can amplify noise. This is usually a high Q circuit. Your well-shielded transformer, when connected into the circuit, may increase the system noise.

A line frequency isolation transformer usually does not protect for over voltage, sags, harmonics, or radio frequency interference. However, ferro-resonant transformers, which can be built like isolation transformers, can protect against transient overvoltage and some sags, and by using multi-phase transformer windings in three-phase transformers, you can reduce harmonics caused by rectification of the secondary voltage. The shields can either help radio frequency interference or make it worse.

Isolation transformers are expensive compared to conventional transformers. Samuel M. Goldwasser on his website has a low cost way of making an isolation transformer, he recommends using two similar transformers back-to-back with the secondaries tied together.

For further reading see Morrison, R., Grounding and Shielding Techniques in Instrumentation, 3rd Edition, Wiley, 1986. This is the best book I know of on the subject and contains both a theoretical explanation and much practical advice on actually building shields into transformers. I had a tough time understanding shields and grounds until I studied this book. It gave me the tools I needed to solve many noise problems in systems powered by switching-mode power supplies.

As a side comment, I never use a shield in a switching-mode power supply transformer. I always found that the shield and its connection always generated more noise than the shield attenuated.

Comment by Ray Ridley, 09-05-2000:

This one I'm not so sure is clear cut. The medical guys use double shielded - they simply have to because they can't use common mode caps.

I think its a function of where (and how) you tie the shields. I don't tie them to chassis ground. As you say, that increases the noise, often dramatically.

I have had success tying them back to the point where the noise is generated, i.e. keeping the RF loop very tight.

[Ray Ridley, Ridley Engineering, has been a sponsor of this website - JF]

Others may have a different perspective on this topic. Comments are always welcome.

Posted by Jerrold Foutz at September 5, 2000 12:56 PM