Hi Robert,
Sorry, I was in somewhere without internet access yesterday.
A newer PCB is not needed since we can use the current one (Rev 2.1) to do the fine tune.
BTW, thank you for your order.
Panson
Nice...! Just remember (and I know you will) to track the bias voltage over a longer period of time (60-120 minutes at least). Thermal systems have a nasty habit of having very, very low bandwidth and may oscillate at sub-Hz frequencies.
~Tom
Panson,
Is it conventional to measure across only 1 Re?
I was thinking that the target 22mv was from the Audioman54 findings - where he found that a bias of about 22mv was better slightly than the "normal" 44mv for the 49811.
I've been looking at my temp simulations and found that the 3k/2k R3/R2 combination give the lowest bias rise with temperature, of about 2x with the Cordell configuration. Using the aforementioned Audioman54 findings, setting the initial bias at 22mV and knowing that it would rise to 44mV under heavy loads the amp should preform pretty well.
I'll try to get the heat sink hotter. I'm only running 33v rails into 8 ohms and it just doesn't want to get much past 40c - unless I bias it into class A - but that goofs up the whole test.
Ken
Ken,
Taking one Re volt drop is fine unless you have a very unbalanced upper and lower half.
Yes, Mark has recommended 22 mA for 49811. My measurements agree with him. You can get the heatsink hotter by using a 4 ohm load.
My target is to have <= 10 % bias variation.
Panson
Panson,
Let me ask another my question another way is Mark's 22mA bias recommendation measured across both Re's - from emitter to emitter?
I have to buy more power resistors...
Good luck on the 10%, I await your solution!
Ken
with the conventional temp comp systems I have used, I prefer to see a small fall in bias current with increasing temperature.
Expect some bias changes from cold to warm. then after it is warmed up, try artificailly heating the heatsink, hairdryer maybe. Or what is the bias immediately after a few minutes at say 30% maximum power?
The overcomp allows for higher bias when the amp is cool and in cool ambient. As the ambient rises the bias falls very slightly. This is the opposite of thermal runaway.
Expect some bias changes from cold to warm. then after it is warmed up, try artificailly heating the heatsink, hairdryer maybe. Or what is the bias immediately after a few minutes at say 30% maximum power?
The overcomp allows for higher bias when the amp is cool and in cool ambient. As the ambient rises the bias falls very slightly. This is the opposite of thermal runaway.
Ken
Is there any difference? Both Re (output NPN and PNP) pass the same current.
Panson
Silly me, I've never compared, I just assumed that the current would be additive. I'll go and compare.
Ken
The output bias is set when the input has near zero Rs and the output is open circuit.
If the output is open circuit and one has just one pair of output devices with one pair of emitter resistors then the upper Ire = lower Ire +- Infb.
Usually Infb is tiny in comparison to Ire. Effectively Lower Ire= Upper Ire.
with multiple output pairs.
Sum of Upper Ire = Sum of Lower Ire.
If all upper Re are matched and all lower Re are the same, then the
Total Upper Vre = Total Lower Vre.
The variation in Vre of the matched resistors is a direct indication of how well or how badly the transistors were matched @ Ibias
If the output is open circuit and one has just one pair of output devices with one pair of emitter resistors then the upper Ire = lower Ire +- Infb.
Usually Infb is tiny in comparison to Ire. Effectively Lower Ire= Upper Ire.
with multiple output pairs.
Sum of Upper Ire = Sum of Lower Ire.
If all upper Re are matched and all lower Re are the same, then the
Total Upper Vre = Total Lower Vre.
The variation in Vre of the matched resistors is a direct indication of how well or how badly the transistors were matched @ Ibias
Yup, half the voltage across half the resistance still equals the same current as measuring across both Re resistors
Changed R3 to 1k and R2 to 5k to try and increase the sensitivity further.
measured the following ( it was cool this morning hence the low starting temp) haven't used Andrew's hairdryer trick yet.
13.8C = 9mv = 22ma
16c = 9mv =22ma
19c = 11mv = 27.5ma
23c = 12mv = 30ma
27c = 12.5mv = 31.25ma
32c = 15mv = 37.5ma
32c-14c = 18c
37.5ma - 22ma = 15.5ma
1.16ma rise per 1degC
Ken
Yup, half the voltage across half the resistance still equals the same current as measuring across both Re resistors
Changed R3 to 1k and R2 to 5k to try and increase the sensitivity further.
measured the following ( it was cool this morning hence the low starting temp) haven't used Andrew's hairdryer trick yet.
13.8C = 9mv = 22ma
16c = 9mv =22ma
19c = 11mv = 27.5ma
23c = 12mv = 30ma
27c = 12.5mv = 31.25ma
32c = 15mv = 37.5ma
32c-14c = 18c
37.5ma - 22ma = 15.5ma
1.16ma rise per 1degC
Ken
oops,
that should be 15.5ma/18c = 0.86ma rise per 1 deg C
Hi Andrew,
I was shooting for 22 to 24mA compensation as recommended by AudioMan54. I can trim the compensation up or down without difficulty. I think the "drift" is pretty constant.
16c = 9mv =22ma
23c = 12mv = 30ma
7c/8ma = .875 degC/1mA
27c = 12.5mv = 31.25ma
32c = 15mv = 37.5ma
5c/6.25ma = .8 degC/1mA
This is about 10% variation, I the results are from my measurement technique - had held infra-red temperature sensor, 1mv DMM range.
I have run out of "tricks" with this design.
Ken
