Bob Cordell's Power amplifier book

I remember you said you use lower than 26mV. I never quite get that. What is the reason you don't follow Oliver's condition of 26mV?

I've seen the "glitch" at < 30ma , Sanken's will be good at 50ma (OEM
recommendation) . the ON's , since I'm runnin' so cool , I went to Jwilhelm's
15-16mv setting.

I just had a open loop condition on one of my amps. The feedback wire
broke inside of a piece of heat-shrink. Spooky went to either rail -
Protection flashed "hard" (shut off speaker) - that "21'st century" just paid for itself again !!
Either blue led cascode reference would be on or off if I touched the PCB!
The LED's told me the problem. The input stage CCS red reds would stay on
regardless.
So the combination of red/blues told me exactly what was going on
with this fault.
Fixed the wire , neither the OPS or IPS went up in smoke !! :eek:
Over designing these modules seems to have worked out.
Very strange I did not smoke this IPS going open loop ? (I've done this before).
Perhaps it is the conservative aspect of the design ? (below)

Sucks to have to troubleshoot something that works so well. :(
OS
 

Attachments

  • spook.jpg
    spook.jpg
    166 KB · Views: 267
I just finished testing the beta of all the MJWxxxx. The 3281 are very good, I can easily get sets of 5 to match Vbe to 1mV and beta to 1%.

The PNP 1302 varies a lot. 4 only have beta from 96 to 101, quite a few from 115 to 120 and a lot from 125 to 131. That's disappointing.

Here is the detailhttp://www.diyaudio.com/forums/solid-state/275364-mjl3281-vbe-matching-2.html

I think I can easily choose sets of 5 that has beta match to 3% and Vbe to 1mV. You think this is good enough to avoid hogging?

Also, I have all 10 output pairs in a roll, I think it will be best to put the transistors that has Vbe maybe 2mV to 4mV less at the end of the roll and have the ones in the middle 2mV or 4mV more. the reason is the transistors at the two ends does not get as hot, so this is to assume the temperature of the two transistor is about 2 deg C lower than the ones in the middle.
 
I think I understand the mechanism of current hogging. The problem is not even at idle, it's during large signal where output has to source or sink a few amps of current. The problem seems mostly because of the value of the base stop resistor of the power transistor because at high output current, the base current is significant that cause voltage drop across the base stop resistor become significant. When there is a beta mismatch, it can cause significant difference in current difference between the two transistors.

Let's just use an example of base stop resistor is 2.2ohm. Let's look at two transistors, one with beta = 100 and the other beta = 120.

1) At 3A output current, the base current of the transistor that has beta=100 is 3/100=30mA. Voltage drop across the 2.2ohm base stop resistor is 0.03X2.2=66mV.

2) At 3A output current, the base current of the transistor that has beta=120 is 3/120=25mA. Voltage drop across the 2.2ohm base stop resistor is 0.025X2.2=55mV.

As you can see, at 3A each, the difference just because of the 2.2ohm is 11mV. This difference will be reflected onto the emitter voltage of the two transistors. Under this condition, the one with lower beta, the voltage across the Vbe and the Re is 11mV less. If the re is low, this 11mV as a lot more effect on the Vbe which control the current. Bottom line, the transistor with beta=120 will conduct a more current, in turn the temperature rise and cause -2mV/deg C and further increase the current.

So my conclusion are:

A) hogging main due to beta difference between parallel transistors.

B) The larger the value of the base stop resistor, the more the difference in the voltage drop across the base stop and more unbalance between the transistors. So you need to keep the base stop as small as the amp can be stable. For example, if you double the Re to 4.4ohm, the difference of voltage drop with double to 22mV if everything else are equal. If you use 10ohm base stop resistor, the difference in voltage will be like 50mV, that's a lot of difference as Vbe is logarithmic, 50mV will create big difference in current between the two transistors. This all caused by beta mismatch of less than 20%.

C) the Vbe matching is only secondary as the major mismatch is caused by the beta mismatch at large signal.

Please let me know whether I got this right.
 
Last edited:
AX tech editor
Joined 2002
Paid Member
Hi Jan
Perhaps it is not wise to contradict my publisher;) but I think Keantoken is correct on this.
Surprised no one has mentioned that this is all covered quite well in Bob's book p299-301.
It is also relevant to Alan's concern's about low Re thermal stability.
The calculation there can determine if his proposals can be thermally stabilized.
My offhand approximations have it as achievable, maybe too close for peace of mind.
But best to have Alan do the numbers and satisfy himself.

Best wishes
David

No disagreement here David - my goal was to explain that current hogging/Re and supply voltage are not necessarily related. The notion that you can prevent current hogging by limiting the supply voltage can give you a false feeling of safety.

That's why I asked Alan to do the numbers and see for himself. And note that in my example, that would clarify the issue for him, does not figure supply voltage in any way.

Edit: I see Alan still doesn't grasp it. I'll let you have a shot at explaining it ;-)
He also is to close to my sig line for comfort -

Jan
 
Last edited:
Edit: I see Alan still doesn't grasp it. I'll let you have a shot at explaining it ;-)
He also is to close to my sig line for comfort -

Jan

Did you see my post above? that current hogging has a lot to do with mismatch of beta and base stop resistor. Larger Re can help equalize the difference in the voltage, but the beta and base stop seems to be very important.

what is sig line?
 
I would not use 0.1 ohm RE in ANY output stage - even a single. It is just my opinion, but I strongly disagree with Self even showing a 0.1 ohm RE example unless he accompanies it with well-grounded explanations about the dangers of doing it. Going down to 0.1 ohm just to get presumably lower distortion is just plain foolish, and is an example of sacrificing other amplifier characteristics for just a better distortion number.

I like low distortion and I like low noise, but I do not believe in sacrificing other amplifier characteristics just to achieve low numbers. Similarly, if you want lower HF distortion numbers by a factor of about 2, just increase the ULGF by an octave - you may get away with the reduced gain and phase margins, but it may not be wise.

I think Self explains the theory of where the 26mV magic number comes from, but I do not think that he gives the credit to Oliver.

Cheers,
Bob

Hi Bob,
Do you have any bad experience with 0.1R in real built amp? To be correct external Re resistor is not alone, it should be added BJT internal Re (about 20mohm if I remember correctly) and the base stopper devided by the beta ( more about 20 mohm or more). That will give about 0.14R or close. I have seen practicalamps with 0.1R for Re, I use 0.16R in mine ThermalTrak amp and no problem what so ever,and Iuse it for more than three years. I did some beta matching, but not to the great precition.
Damir

PS. I would liketo see in new edition something about current conveyor use as the voltage gain. It is very low distortion gainstage even with no global feedback and is very well suited for the CFA.
 
Last edited:
Surprised no one has mentioned that this is all covered quite well in Bob's book p299-301.
It is also relevant to Alan's concern's about low Re thermal stability.
The calculation there can determine if his proposals can be thermally stabilized.
My offhand approximations have it as achievable, maybe too close for peace of mind.
But best to have Alan do the numbers and satisfy himself.
Sorry I missed your post as you did not address to me directly. I think this part support my idea of lowering the rail voltage and that's the first element of Fig. 14.15. Rail voltage Beta_T1 is a main cause on this feedback. Beta_T2, Beta_T3 and Beta_T4 are the heatsink and the result of Beta_T1. Given the heatsink and transistor package(you can only choose the chassis and transistors), reducing rail is the key.

No disagreement here David - my goal was to explain that current hogging/Re and supply voltage are not necessarily related. The notion that you can prevent current hogging by limiting the supply voltage can give you a false feeling of safety.

That's why I asked Alan to do the numbers and see for himself. And note that in my example, that would clarify the issue for him, does not figure supply voltage in any way.

I did the numbers as shown in the last post to show Re is not the only one that cause the problem. Base stop resistor is actually a very big part of it. This is confirmed in page 302 of Mr. Cordell's book. If people use 4.7 or 10ohm base stop, this can be very serious and cause hogging.

I am not trying to say I want to go back to 0.12ohm or 0.1ohm. I just say Self keep saying I found 0.1ohm is good enough for all condition over and over. I just bring it up to see what you guys say. I did not put the 0.12 back to the board and I don't think I will go this low. I might consider 0.18ohm which I ordered to compromise.

If you have other ideas, can you explain?
 
Last edited:
I did the calculation in p300 to 301. And I did the 1/3 power output also. That really shows the limitation using 0.12ohm. I can only use up to 25V rail even though according to calculation in p300 shows it's good for 50V. Here is my calculation, please comment:

I want to find the max rail voltage and still get Beta_T <=0.5. So GT=1/(1-0.6)=2.5. I want to calculate the max rail voltage to stay in safe area.

For static bias of 200mA and 0.12ohm resistor
1) gm=1/(0.13+0.12)=4A/V.
2) Thermal resistance from junction to heat sink = 1.1deg/W for TO-264.
3) TC_Vbe =-2.2mV/deg C.
4) Vrail = 0.5/(4A/V X -0.0022V/deg X 1.1deg/W)=51.65V.

So if I have rail voltag less than 50V, I should be OK ONLY in static wise.


But for 1/3 power test, average current is over 2A, we can assume r'e=0. So we use Re to calculate gm= 1/0.12=8.333.

Vrail = 0.5/(8.333A/V X -0.0022V/deg X 1.1deg/W)=24.8V

This mean I cannot safely use rail voltage over 25V.

From this calculation, Self cannot be right that it's safe to use 0.1ohm. It's very limiting to use 25V. I did the calculation, with 25V rail and 5 stages of 200mA per stage. I get 8W class A, but the max power into 4ohm load is only 60W, worst is 30W for 8ohm load.

This is still assume matching transistor. The post I calculate the effect of base stop resistor is still valid, in fact the effect is quite huge if the beta of the transistors are not matched. It just goes down hill.

I did calculation with 0.15ohm, max rail voltage is 31V. For 0.18ohm max rail voltage is 37V. I have to use 0.2ohm in order to use 40V rail.

From this, I think you guys and Mr. Cordell is correct not to use anything less than 0.22ohm. And at that, from my calculation tell me you cannot use over 46V rail!!! Maybe I can up the Beta_T to over 0.5, but you cut into the margin.
 
Last edited:
Hi Mr. Cordell

Can you take a look at post #5924 and comment on of my calculation on the error caused by the voltage drop across the base stop resistor due to beta mismatch. In idle condition with only smaller bias current, it's not important. But when I calculate with medium signal, 11mV error between two stages one with beta=100 and the other 120 with low base stop of 2.2ohm.

It goes down hill fast if people use 4.7ohm or 10ohm!!! Is my number wrong? We are talking about 22mV difference between the two stages with 4.7ohm. The log nature of Vbe will cause big current mismatch between the two transistors.
 
I kept thinking, there must be a factor of 2 I missed. In dynamic power mode, the transistor is off half the time due to Class AB, so there have to be a factor that help, giving safe margin to higher rail voltage. This is unlike the steady state where it's on 100% of the time.

Too many people having success with 0.22ohm and 80V+ rail.

We need Mr. Cordell to look at it.
 
From this, I think you guys and Mr. Cordell is correct not to use anything less than 0.22ohm. And at that, from my calculation tell me you cannot use over 46V rail!!! Maybe I can up the Beta_T to over 0.5, but you cut into the margin.

Even if it is thermally stable, current hogging can still occur. It might just stop with one transistor having twice the current of the others for instance. You want to have a large margin from the unstable point.
 
Even if it is thermally stable, current hogging can still occur. It might just stop with one transistor having twice the current of the others for instance. You want to have a large margin from the unstable point.


This is calculation of a single transistor regardless of hogging between parallel transistors. I was confused by it too until I read the first line on page 300 that it said "Remember, this analysis is for one output transistor with the temperature of the heat sink constant in the time frame of the analysis."

This is for single transistor without any help from other parallel device, more transistor in parallel can only help even though you have current hogging. This is the worst case.

I think I still miss the factor of 2 as I calculated as if the transistor is on 100%. In class AB, the transistor is only on somewhere over 50%. So it's not as bad. But I think Self still is not correct. He did not put the limit on the rail voltage. Even if best case where I missed the factor of 2,you cannot exceed 2 X 25V =50V rail before it gone poof. Like you said, other factor can easily comes in and you never want to push to the limit of the calculation. Bottom line, 0.1ohm is truly dangerous.
 
Last edited:
Did you see my post above? that current hogging has a lot to do with mismatch of beta and base stop resistor. Larger Re can help equalize the difference in the voltage, but the beta and base stop seems to be very important.

I did not match - (below) is the 72V readings on the worst of the two amps.
The spread of the readings was about 1/2 (under 1.5mv) at my test supply
(45V).

.22Re - about the same cold to after heavy (I mean real heavy ! ) use.

OS
 

Attachments

  • slew200.jpg
    slew200.jpg
    123.6 KB · Views: 198