Measuring O/P BIAS current

Hi All. I just cannot get my head around this BIAS measuring of the L12 Amp to enable to set the bias trimmer! QUESTION: What VOLTAGE should I set across the combined emitter resistors please in the shown example below to achieve a bias current of 40mA? It's a BASIC OHMS Law calculation, but do I use the SUM of the resistors (0.44 Ohms) or just the single value of 0.22 Ohms?. ... In my amp the emitter resistors are in fact just 0.1 Ohms.... Perhaps I should measure each collector voltage separately to ground across it's single resistor instead? Many thanks for your help.

1749304511069.png
 
Hi,
With input shorted, volume minimum and no load connected ... Is that total; output stage bias current, or each device? For total:

Measure between your two test points and calculate as voltage over 0.44 ohms. If you could accurately measure the resistance, do that. You would need a Kelvin, or 4 wire resistance measurement to do this.

Your meter must be accurate at that voltage, so a 3 1/2 digit meter with 100 mV scale will not cut it. Your target is 17.6 mV.

For each device, measure across the 0.22 ohm resistors in each emitter circuit. Your target is 8.8 mV, demanding a good meter. The voltage drops should be close to each other, indicating a good match. I would check this even if the total current was measured across the points you indicated in your diagram, but your target voltage for total is only 4.4 mV in that case.

Once you have it set, confirm your AC mains supply voltage is what it should be. Let it run for 10 minutes or so and check it again. Then if you haven't needed to correct the adjustment, connect a load (assuming DC offset is okay) and play some loud music into a dummy load for 10 ~ 15 minutes. Disconnect the music and load and measure the bias current again. Allow it to settle until it cools off and measure again.
 
The voltage across each resistor is just the current through that resistor times the resistance of that resistor. As the polarities are the same, the voltage between your arrows is the sum of the voltages across the two resistors.

If the currents through them are the same, which should be the case to a good approximation when the output of the amplifier is left open, the sum of the voltages across the resistors is just the current times the sum of their resistances:

I R13 + I R20 = I (R13 + R20)
 
Very many thanks anatech and MarcelvdG. Very much appreciate your comprehensive replies. The voltage of 17.6mV is what I calculated to achieve the 40mA bias current. What was throwing me was the YouTube video by Mike Beeny, that shows just 8.8mV being set across the collectors (as per my red dot positions) . I thought this must be wrong but then I thought that as the junction between them is zero volts (ground) I might be wrong. Just as a final check for my amp with 0.1 ohm emitter resistors please, I should aim to get 8.8mV across the same test points (half of that for the 0.22 Ohm resistors)? I attach the original circuit shown for the bias mod, (note the Bias figures and typo error)
1749312449119.png
 
When R13 = 0.1 ohm and R20 = 0.1 ohm and you still want 40 mA of current, the voltage across each should be 4 mV and the voltage between your red arrows should be 8 mV.

I'm assuming here and in post #3 that the 40 mA bias current you wrote about is the target for the entire CFP/Sziklai output stage, so on the positive side, the sum of the currents of Q5, Q2 and Q10. Is that correct?
 
Thanks for the correction MarcelvdG..... Yes I have 0.1 Ohm resistors and the sum is just 0.2 Ohms for the pair, so just 8mV is the required voltage then....Thanks..... I am only following the mod as originally shown on DIY Audio some time back. It would seem the figure shown was the TOTAL of the entire output stage, so the answer is yes. I should really set the bias up using an oscilloscope, as others have done. It seems as standard, there is some cross-over distortion. For reference, the standard fixed 1K ohm resistor set the total bias current at around 25mA. There is a temptation to use even more bias but I understand this could actually result in more distortion (Douglas Self)
 
I'm not sure of your oscilloscope method. What I do is run a 10KHz signal into a 4 ohm load. The output is monitored with a distortion analyzer and one channel of an oscilloscope (triggered on this waveform). The monitor output of the THD meter is monitored by the other channel of the oscilloscope and you can then very clearly see the crossover notch.

Increase bias current while watching your bias current level. This is merely a safety and reality check. Some designs will continually lower the crossover notch with increasing current, while a good design (in my personal estimation) will drop the notch below the noise floor at low bias currents, lower then specified.
 
  • Like
Reactions: sonicles
Hi anatech. I managed to set the bias yesterday at around 8.3mV over the 2x 0.1 Ohm emitter resistors (as standard around 4mV). These were a new pair of boards. When I switched on I found the very bright sound I did not like was still there (to my disappointment) however after around 6 hours of use I think the sound has settled down (mellowed) and I am starting to appreciate the sound and now I actually think it has a lot of merit.....certainly "fast", extremely dynamic, with a tuneful, powerful bass. Whether I have the bias at optimum, I just don't know. If I had the test gear, I would do as you suggest. It runs cool (It has very big heatsinks) and also does not drain the twin power supplies very fast (takes several minutes as the PSU caps charge is indicated by 2x leds per channel and they stay bright for ages) With this apparent low quiescent drainage, I think I could increase the bias say another 10mA and just see if the amp sounds even better. I can only think the original extreme brightness was that it needed "burning in". I was always sceptical about this phenomenon but I'm sure it's not just my ears "getting used to the sound" (although some could be this). I have to say it would be nice to really know if I do have the absolute optimum bias distortion wise. Maybe when I get more time I will arrange some testing with a 'scope and generator (square waves) Thanks again 🙂 PS: This L12 amp really does prove that amps CAN sound so very different, if one was to guess at the difference in frequency response to the L20, it is like someone has turned the "treble control" (I don't have one) up by several dBs.
 
Last edited:
Hi sonicles,
There are other reasons why something might sound bright, or unpleasant. These reasons have nothing to do with output stage bias. Crossover distortion can sound unpleasant, but that is it's own thing.

Playing with an amplifier ... you really do need the test equipment and training. There are many people with these things that still have zero idea of how even a simple amplifier really works. So while you can fix an amplifier without the training and tools, it won't be repaired and running properly except by chance. You also stand an excellent chance of causing immediate or future failure.

About measuring. 8 mV is a stretch for most inexpensive meters. I know hobbyists are not trained metrologists. Most technicians don't really understand either. Here is an exercise you should try. Look up the accuracy specifications of your meter. Apply all the errors and figure out with the range of indicated values might be for that particular measurement. Look this up to do the calculation properly. It's pretty dry stuff, sorry. Anyway, it will tell you what your meter (tool) is capable of and whether you can trust it.

Now to really dampen your spirits, many inexpensive meters are out of tolerance, new out of the box. In addition, accuracy becomes worse with time and also the difference for the temperature the meter was calibrated at. If your meter doesn't have specifications for each function and range, it is probably because they are pretty bad. You shouldn't trust that meter for anything but checking to see if mains voltage, or DC voltage is present and a very rough resistance reading. Current, resistance and AC readings are always worse than the DC specifications. You'll also see the last digit (LSD) is often meaningless.
 
  • Like
Reactions: sonicles
About measuring. 8 mV is a stretch for most inexpensive meters.
Hello anatech. I think my meters are accurate to about 5%. I was not worried as to get a fixed precise amount of bias, getting the 2x sides the same was just as important to me. At the moment, after rechecking I notice the long term voltage has crept up to something like 9.3mV.... I also measured the actual trimmer resistance value, and was amazed I had them both adjusted to exactly 1006 Ohms each! I found this reassuring that this precision had been achieved. This equates to around 47mA, still a very low figure and the amp runs only a few degrees above room temperature and doesn't warm up much at high power, long term usage. It is now sounding quite smooth, but yes, I would like at sometime to bench test the distortion and tweak for the lowest amount. ...... DC offset at the O/Ps is 40 and 47mV. This figure has not changed and is around the same for the other pair of boards I have. This would be caused by some component value variation in the front stages. I think the amount is acceptable as it causes no click in the speakers at all when the relays connect up
 
Hi sonicles,
Your DC offset is a reflection of the match in your input devices. Some circuits can have substantial DC offset when everything is perfectly matched, but this should be much lower.

Transistors must be matched when they are at exactly the same temperature. In use, they should be in thermal contact, some isolation from ambient is helpful too. Decades ago I designed a DIY transistor matcher and gave it to this community. it is simple and works extremely well. It is basically a diff pair circuit with CCS tail current (selectable). You can't measure beta separately and end up with really good matches except by chance. This would be better than just picking transistors out of a bin and using them, but you can do much better.
 
... and that is if it is in tolerance to begin with, if it even was when new.

Most meters costing less than $200 CDN were not. Not only that, due to the input divider type used, they could never be adjusted to be in tolerance.
 
My Topcraft TDM 600 digital multimeter that cost me 10 euro or so is specified to have an accuracy of +/- 0.5 % of the read-out value +/- 2 digits when in the 200 mV DC range. At 8 mV, that boils down to +/- 0.24 mV. Not too bad, considering that the quiescent current will vary with programme dynamics anyway.
 
  • Like
Reactions: sonicles
Hi Marcel,
Forgive me, but that is only if that meter is in tolerance.

Also, that will be error at full scale, not the indicated value. So 1mV, +/- 2 digits plus any other error terms. A far cry from 0.24mV absolute. Temperature and time from last calibration factors in as well. Accuracy specs are loose and sparse, so expect the worse and then some.

That is the DT-830D, a really cheap meter. I would only trust it to tell me if I have "about" 12 VDC, or 120 VAC in a circuit. It is rebranded and very popular, but absolutely not something you can trust. I actually have one!
 
The spec says quite clearly in five languages, two of which I understand, that it is a percentage of the read-out value plus 2 digits. Full-scale is 1999 digits, so if it were +/- 0.5 % of full scale +/- 2 digits, they would simply have written +/- 12 digits.

It's essentially an ADC with a display, so it makes sense that it has an offset, a gain error, a round-off error and a bit extra due to DNL and INL. In practice, with the input shorted, it always indicates 0.0 mV, so at least the offset is well in spec.
 
That usually is 5% of full scale. So on the 100mV scale, the accuracy is +/-5mV.
The consequence of that when measuring a nominal 10mV should be obvious .
Hello Jan, I answered incorrectly before My meter is a 4 1/2 digit +/-0.5%+3 type....not 5% FSD! It is good enough to set the bias that does not have to be 100% accurate to a fixed figure. One objective is get identical currents on both channels by careful adjustment
 
  • Like
Reactions: jan.didden
Generally speaking, these meters are adjusted on one scale only. Typically 1.9VDC on a 2V scale (or 190 mV for 200 mV). Whatever the full scale voltage is, normally 95% of that. Your zero is normally auto adjusted on dual slope and other converters. If it didn't read zero, you have serious problems. Other scales typically have no adjustment, are not correctable without component replacement. Everything depends on the voltage divider, and if is isn't a deposited film type on ceramic substrate, laser trimmed, it will not track for temperature or have decent higher frequency characteristics. Then you have your AC conversion circuit, and current source for resistance. The current sense resistor affects that function. Everything hinges on the DC voltmeter circuit. It will have the best accuracy specs.

Less expensive meters will be shy on details and probably will not have temperature or time from calibration specs. You can't trust those meters without checking against a meter with known specs. the other issue as I mentioned earlier is that the cheap meters are not calibrated carefully, and the factory standard may also be out of tolerance. Every make wants to have "good specs" and some are outright lies. You really need to make sure, because cheap meters are sold to make a profit only. Many are contract manufacturing with no one responsible for problems. Does this sound like you can trust them?

If you need to trust a meter, Keysight or Fluke. Some others can be good, like the Escort brand was, maybe a couple others. None are inexpensive, adjustment costs money, as does proper design and manufacture.

They call it "fat, dumb and happy" for a reason.
 
  • Like
Reactions: sonicles