Distortion at about 6-8K.

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I have built Rod Elliots project 72 gainclone, but have substituted a opa544 for the LM1875. After a fue teething troubles everything seemed ok.
However after extended listening there appears to be heavy distortion at about 6-8K (my guess). This sounds like typical cone breakup distortion, but seems to clean up above and below the distortion band. I checked this on three different sets of speakers (all single driver fullrange) and the result is consistant.
The only major departure I made from the original design spec is to use a 2.2ohm resistor in place of the 1ohm resistor in the Zobel. I find it hard to believe that it is this though as a single fullrange driver is a very easy load to drive.
The circuit has high frequency input filtering and the zobel so it should be well protected against ocillation. The opa544 shows higher THD figures than the LM1875, but I would imagine that most of this is harmless second order as the design has a FET input stage.
Could it be a difference in input impedance between the two chips.

I am fishing here guys so help me out.

Shoog
 
It has a large EI transformer (at least 180VA), which is showing no signs of strain.
It certainly doesn't sound like the sort of clipping I heard with the LM1875, so I would say it isn't clipping.
The heatsink on the chips is not even above ambient.
Oscillating would be my guess, but until I can present it to a scope again I don't know for certain.

Shoog
 
I tracked down the source of the problem. It was output stage instability. I have put a 10ohm resistor in series with the output, and hay presto the distortion has completely disappeared.
The sound is no great. Detailed, tight bass, silky smooth. I would tend to concur with Pete Daniels when he says that he cant understand why he went to all the trouble of building Alephs when the gainclone does it just as well.

Shoog
 
Putting a 10R resitor in series with the output will degrade you sound quality, better to use a small heavy guage wire inductor often put in parallel with a 10R resistor (or even wound round it if it is s non ferrous resistor). This will normally have the same stabalising effect without ruining the damping factor of the power amplifer in the audio band.

Andrew.
 
I tried that and unfortunately the distortion returned.
It could just be that my speakers which are vintage valve design just dont like tight control. Putting in a resistor to make the amp behave more like a valve amp might be just what they need.
It turns out that I designed all my speakers when I had a valve Leak stereo 20 and so they are all tuned for valve amps. I even had to tune my Zen V3 to sound more like a valve amp, to match my speakers.

Maybe that is all the distortion is, but I have my doubts. It could be that the Zobel components are super critical and I absolutely must use the 1ohm resistor as specified. At the moment the result is good enough and I will probably leave it.

I might try using my last remaining OPA544 to build one of the ultra striped down gianclones I have seen mentioned here, and try using it with a pair of more conventional speakers.

it has however sujested a fix for a pair of speakers (vintage alnico siemens 4inch drivers in Buscchorn MkII) which I am tuning for a friend who will be using a musical fidelity B1 Mosfet amp. I will try putting a 4.7ohm resistor in series to fool the speakers into thinking they are looking at the valve amp they were made for.

Thanks for the help again

Shoog
 
Introduced 0.47ohm resistors in series with the outputs. Cleaned up the distortion somewhat, but still there.

Had it on the scope today and saw the same distortion component across the frequency range, apart from below about 1K. On the rise to the sine wave there is a little kink just before the top. On the fall of the sine wave there is a matching kink just before the bottom. This sounds like either speaker basket reasonance or cone break up.
Thought that it was crossover distortion at first, but it doesn't appear at the zero point on either side.

Anyone seen this type of distortion before, anyone care to sujest how I might get rid of it.

Shoog
 
I believe I have worked out what the distortion is. Its current limiting that is dumped to earth (hence no massive voltage spikes).
I went back to the spec sheet and found the fundamental difference between the LM1875 and the OPA544. The OPA544 has a maximum input sensativity of 0.7V, so the two volts output of a CD send it into automatic current limiting and hence nasty distortion.

In order to get thing working I changed the input resistors for a voltage divider, with 22k on the input, referenced to a 10k to ground and finally followed by a 2.2k at the input of the opamp. This has finally sorted out most of distortion- need more observation.
I am probaby wrong in my implementation of this mod (how it effects source and load resistances) but at least the thing is working for the first time.

Any sujestions of a better set of resistor values to reduce the line in voltage

Shoog
 
In the end it turned out that the OPA544 is totally unsuitable for use as a gainclone. This distortion is a product of its design and is a result of not been designed to drive low loads. My speakers where 4ohm so they were overloading the chip. The nature of the distortion was simple crossover distortion leading to various complex intermodulation distortion components.
If you are trying to get the OPA544 to work as a gainclone - don't waste your time.

Shoog
 
Not really then. The LM1875 is an extreamly sweet sounding little chip (it was my first gainclone). I would start by looking at all solder joints. Could be a minute solder splash loading one of the rails to ground. Could be a faulty chip. distortion in this range suggests crossover distortion - which suggests an overloaded chip.
Have you scoped it out. Have you tried it on a different load/speaker. ie are you absolutely certain that its the gainclone thats at fault.

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