Bob Cordell's Power amplifier book

Jump to Latest
B

Bob Cordell

Member
Joined 2006
cordellaudio.com
mchambin said:
Distortion minimized at what output signal level ?
There are as many Oliver optimums as output level choices.
Click to expand...

You are exactly right, and therein lies the matter of compromise. THD vs output level for several different values of bias needs to be looked at and then a compromise bias current level needs to be chosen. It is usually the case that THD will be at a maximum at a certain broad power level for a given value of bias. That broad range of power level where crossover distortion is highest will shift a bit for different bias levels. For a 100W/8-ohm amplifier, the range where crossover distortion is highest is often below 10 watts.

In examining a given design for what bias level to choose, I will often measure just the output stage distortion open-loop at 1 kHz. By this I mean that I close the feedback loop prior to the output stage, usually from a center-tap on the pre-driver emitter resistor in a Locanthi T circuit output stage. This makes the output stage distortion easier to see by itself, and of a higher percentage level, and also makes it possible to do the measurement at 1 kHz without there being a huge amount of frequency-dependent negative feedback around the output stage.

In the absence of doing it this way, I will often just look at THD-20 as a function of bias level at some chosen low to medium power level.

If, based on your measurements, you are unsure of what bias level to choose, I recommend that you always err on the high side. You never want bias tracking errors to cause the class AB output stage to go into a serious under-bias condition. If you want to look at these sorts of things with a given bias setting, first measure the quiescent bias voltage across the emitter resistor after the amplifier has fully warmed up and stabilized at zero power output. Then run the amplifier at 1/3 rated power into 8 ohms for awhile. Then kill the audio signal and observe the bias voltage as a function of time. It may move quite a bit.

Cheers,
Bob
 
S

stuartmp

Member
Joined 2016
Paid Member
Bob Cordell said:
Similarly, distortion in my BJT designs is usually overall lowest with about 18-20 mV across the emitter resistor, even for designs with just 2 output pairs. These designs typically have 0.22-ohm emitter resistors and 2.2 ohm base stopper resistors.

Cheers,
Bob
Click to expand...
Hi Bob and everyone else.
Thank you for all that detailed information. It's certainly given me also to read over.

I'm going to try doing some wingspread simulations this weekend.

Just wanted to check with you Bob on your last post. When you say 18-20 mV I am assuming this is across one of the emitter resistor and not both of them?
 
M

mt490

Member
Joined 2010
keantoken said:
The reason the range extends below 26mV is because while 26mV results in a better average distortion when operating above the crossover region, it does this by creating a gain dip at the crossover point, thus creating the W shape in the wingspread chart.

This means that although large-signal distortion has been decreased, small-signal distortion at the crossover point has been increased.
Click to expand...
Another detail in practice is what the drive impedance looks like at the input side of the OPS, of course one aspect that Triples aim to negate the effects of though.
 
Bonsai

Bonsai

www.hifisonix.com
Joined 2003
hifisonix.com
Paid Member
Yes, LSN (Large Signal Distortion) due to VAS loading is often overlooked. The cure in almost all cases is to go for an EF3. I have a physical design which goes from 7 ppm at 100 watts out into 8 ohms to 10 ppm 100 watts out into 2 a Ohms - so for constant power output, virtually no increase in distortion. For the same output voltage swing, the distortion goes to 30 ppm with effectively 4x output power - a good result.
 
B

Bob Cordell

Member
Joined 2006
cordellaudio.com
Book discount

I just noticed that Routledge/Focal, the publisher for my second edition, has the book available for preorder at a 20% discount. This discount does not show on Amazon, where the discount is 5%. You can just go to routledge.com if interested.

I have updated my website at cordellaudio.com to reflect the second edition. The website also includes direct links to the Amazon and Routledge web pages where the book is available. There is also more information on the second edition, such as preface and table of contents, on my site and the Routledge site than on the Amazon page. Availability is still listed as May 31. I am reviewing the proofs as we speak.

Cheers,
Bob
 
R

rsavas

Member
Joined 2012
Hi Bob and others,
I went through the old book, looking for answers about matching transistors for the o/p stage. Found some information, but maybe you can elaborate.
There is some discussion in another thread about matching bjts in a OPS
Matched output transistors: does gain change with age?
Chris was referencing the OnSemi datasheet (MJ21193, MJ21194) where they showed the test of measuring open loop thd in a pair of devices, matched and un-matched.
How much matching needs to be done and how?
Got to wondering if you're matching OPS devices, make any sense to select/match the ballast(emitter) resistors too?
Thinking that the processes are so fine tuned now it might be difficult to find matches if there is no overlap in hfe?
I see ECW20N20-S - Selected Lateral Mosfet selling matched pairs of lateral fets
 
Last edited:
You must log in or register to reply here.
Top Bottom