Balanced Signal Through Entire Amp?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I'm currently working through Kevin Gilmore's electrostatic amp at Headwise's projects. I'm going through and trying to figure out why he did what he did at different places, and I was curious what the DIYAudio crowd had to say about this particular design. Any criticism? Is there a simpler way to reach the goals (electrostatic headphone driver, with +/- 800 volt swings)? Or is Mr. Gilmore's design the best way to go? I'd appreciate anyone's view on this design, I understand the workings of the amp now, but I don't have enough experience to judge the pros and cons of the design.

One particular question I have is regarding the balanced signal path through the entire amp. Is this a normal practice? I've seen balanced inputs of course, but as I've always seen it done, you just pull a signal off one of the sides of the input differential amp, and the rest of the stages are single-ended. What are the advantages of staying with a balanced signal path throughout the amp? Is this merely to provide a +/- swing for the electrostatic panels, or is there some other reason at work here?

I'd appreciate any suggestions / criticisms. Once I perfect an electrostatic headphone amp design (to go with my homebuilt ES headphones), I'd like to scale it up to build a direct-drive solid-state ES loudspeaker amp. So any simplification I can manage for the little guy will help me enormously when it comes to the biggie. Thanks everyone for your help.
- Jonathan
 
Balanced Amps

I have seen many balanced amplifier circuits and very few have any kind of compensation to control common mode rejection, especially at the higher frequencies. Thus many of these designs do not provide optimum performance.

At times the term balanced is used more as a buzz word rather than a true improvement in performance. After all when you nearly double the parts count something has to give. You win some and loose something some where else. Its a trade off.

In special applications it may be benificial to run balanced circuits, the ESL driver being one. For general use however in a well designed system that is not spread over hundreds of feet there is little if any benifit.

John Fassotte
Alaskan Audio
 
Balanced topologies

A well designed balanced circuit does offer advantages over a single ended one. The main one being the rejection of common mode noise on the interconnect for a greater signal to noise ratio. Also the constant load on the power suppy since one side is pushing while the other side is pulling minimizes modulation of the supply by signal currents. And last but maybe most important is that power supply noise is seen a common mode signal and is rejected by a symmetrical and well matched balanced circuit. For output stages you can run lower voltage rails for the same voltage swing allowing higher bias currents for a given power dissapation. Sounds like a hellava deal for some extra parts (maybe not quite twice as many). Go look at the large number audio components offering balanced inputs. Somebody thinks it is worth doing. Even many Dac ICs have balanced outputs. This is out of the subjective camp and well into measurable differences.

H.H.
 
More or Less Balanced

I agree with HarryHaller. But i still say that many balanced circuits actually do not offer maximum common mode rejection due to poor design of the input circuits.

You cannot assume that just because a circuit appears to be balanced that it is. Well balanced designs require components that whose input resistances are matched to better than .01 percent. Also the input capacitance has to be very precisely matched. This usually requires a low frequncy, say 1 Khz balance adjustment pot along with a trimmer capacitor to handle the balance at 100Khz or higher. Without such adjustments the common mode rejection will never be optimum.

When the high end industry needs to make a few more bucks they come up with methods to do this. Like I said balanced circuits have there obvious uses but are not needed in all circuits or for all functions.

Just what is the common mode rejection of typical so called balanced circuits from DC to 1 Mhz. Do you know how far your circuit is from optimum without common mode low and high frequency balance adjustments?

John Fassotte
Alaskan Audio
 
Balanced inprovements

Once again I agree with John. It seems to be the matter of degree that his argument is based on. How about some numbers for what you consider a well balanced circuit in terms of common mode rejection ratios. Also some sugestions on measurement techiques for CMRR would be very interesting. I have even used a quasi balanced input for the SPDIF digital inputs with some very positive audible results!

H.H.
 
Common mode rejection measurement

The simplest way of measuring common mode rejection is to tie both the positive and negative inputs together. The Inject a test signal into this common point and circuit ground.

I typically use a 1 Khz and 100 Khz square wave injection signal. The level of this test signal can be quite high since the only signal that will be amplified by the circuit under test will be the difference in levels reaching the actual balanced input stage devices.

Typically the common mode balance adjustments will be on the positive input for a amplifier that incorporates some negative feedback. A variable resistor is required to adjust low frequency balance and a very small variable capacitor for the high frequency balance. These are adjusted for minimum output level at the output of the amplifier. These adjustments can be fairly critical and typically vary the values of R-in or C-in on the positive input by fractions of a percent.

I would consider good low frequency common mode rejection to be better than 100DB down and if possible I attempt to get around minus 80Db at 1Mhz from say a amplifier that is flat to 1 Mhz or higher. A amplifier that has no common mode adjustments may only be able to reach a minus 40 to 50DB at the low end and be even worse on the high end.

Circuit board or wiring layout is critical to obtain these figures. A finger placed a quarter inch from one of the components on the input of a well-balanced input stage can degrade high frequency balance by a substantial amount. Perhaps in the order of 20 or more DB depending on the impedance of the circuit.

There are many examples of how to measure common mode rejection in the manuals provided by many IC manufactures. I have a hard time understanding why when we go to all the trouble of designing balanced stages that common mode adjustments are ignored. To me it is worth doing things right to get maximum performance. Perhaps I’m just to critical but when I build balanced circuits I expect top notch performance and that requires the ability to adjust input circuit parameters.

John Fassotte
Alaskan Audio
 
John:

Your method does not always work, especially the way "balanced" is made by the most high-end types. There will be error currents that could throw things off. That is why the op-amp manufacturers do things different. Analog Devices has a publication "Best of Analog Dialog" that explains some of this better. And then there is my old buddy Bob Pease (yet another needless plug.......). EDN had an article about noise in "high-performance analog circuits" that also had good stuff on this. Maybe with luck, I can come up with the issue if you like. Most of my stuff is packed away, including the AD book. But I think I know where that copy of EDN is.

You are right about how critical layout is. But I feel asking 80 dB at 1 MHz may be a bit much. Trying to get the passives matched that close, and all the stray capaciatnce will be tough enough. Then you have to take into account the active part of the circuit.......

Look at the CMRR of some good op-amps. I doubt you will see many that high at 1 MHz.

Harry:

Have you ever tried the old ham radio trick of a "sorta-balun"? A cheap way to force the currents to be balanced. If not........you are the pro at making little ferrite thingies.........try it and let us I know. I don't have any suitable cores I can put my hands on right now.


Jocko
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.