I think there is a problem. Although LME offers a current source of 2.45 mA, the diode current is temperature dependent. At high temperature, the diodes pass more current. The amount of drift will change. At 25 C, we have -2mV/C for 100 uA. At, say, 50 C, diode current will increase and we have a different drift.
What do you think?
Hi Lazy CatThis driver/output devices configuration is pretty happy with low Z loads.
An externally hosted image should be here but it was not working when we last tested it.
Have you actually built and listened to this circuit?
Correct Panson, at 50 °C the Vb of diode having changed, the current in the diode is now about 200 microamp this implies ( log 2) a change in drift of 0.3 x 0.2mV/° The drift is now 1.94mV/°C instead of 2mV/°C. Recall that Vb drift of a diode is changing as 0.2mV/°C per decade of current change ( drift increasing in absolute value with decrease of current).
The overall tracking should still be very good.
JPV
Normally the 'theoretical' optimal bias shoould be such that gmRe=1
This is equivalent to IcRe= Vt= kT/q
Therefore Ic ( bias current should be PTAT) so INCREASE with temperature.
This is of course theory but could explain why keeping Ic constant will end up in underbias when transistor becomes hot.
What I am suggesting to Panson is a way to adjust the different drifts of the TT diodes and the BJT transistors. Of course it is not perfect but it is closer than using the diodes as is.
Perhaps the not mached lower drift ( in absolute value) of the diodes is better suited, it should be tried
When temperature is rising, the Vb of the diodes decreases therefore the current in the // resistors decreases therefore the current in the diodes increases ( because current source), therefore the drift of the diodes so of Vbias increases with temperature. It g oes in the same direction as Keeping Ic PTAT, so the change in drift with temperature of the diodes is not a bad thing !!
Because all this is theory, the best would be to play with these aspects and see how the distortion is changing with temeperature and bias/drift settings.
At least, the system gives degrees of freedom in experimenting.
JPV
You guys might get sick of me for bringing this us again, but, back at post 243, I put in a copy of Bob Cordell's bias scheme for the thermal traks. I've built this scheme with the exception of no predrivers. It works as promised, I think I also posted the THD plot in this thread or Panson's other LME thread.
There is more detail in the Thermal Trak thread, search Thermal Trak and DouglasSelf and you'll find the thread.
The actual post was somewhere around December 31, 2008.
Ken
There is more detail in the Thermal Trak thread, search Thermal Trak and DouglasSelf and you'll find the thread.
The actual post was somewhere around December 31, 2008.
Ken
I don't have the sophisticate equipment that you have, but, did a series of THD runs and found that the Thd had very little variation, run to run. The cool thing about the circuit is that you can tune it - if the diodes over or under compensate relative to temp, you change out the resistors until they track perfectly.
Ken
No I do not mach. I would think that mismach of Is the saturation current of NPN and PNP pair will change completely the distortion picture. But we have to start from somewhere to understand the relative merits of each design proposal.
TT transistors allows you to approach the difficult aspect of transient temperature effects on distortion.
It is also well known that in a cascade of amplifier stages, the distortion of the last stage is predominent for weak non linearities. For large non linearities ( slew rate, crossover ) it is different. If the input stages are free of slew induces distortion and crossover distortion, then the main effort should be applied to optimization of the output stage. The use of a LME 498xx is perfect for that and you can concentrate on the main cuprit ( don't forget the 80/20 rule).
Because you are well tooled with your AP, i believe that you can make interesting trials with the triple/LME498xx toplogy and TT diodes.
Trying to measure distortion with different tracking is a very interesting aspect to discover.
JPV
Well I believe it should be investigated practically but applying the 80/20 rule I would prefer to understand the impact on distortion from all the design parameters of a triple with TT transistors and THEN investigate error correction. Thermals are influencing all the topologies, model accuracies, disortion mechanisms.
Do you intend to investigate the distortion with varying currents in the diodes? This should be easy with your AP.
Another interesting test would be to generate an impulse which is a swept sine wave from f1 to f2 in a time T. If T is short but long enough we have a transient whose energy is still well concentrated between f1 and f2. This transient is more or less like a square pulse in the frequency domain ( there is of course energy outside f1 to f2 but way lower if (f2-f1)xT >1)
By applying this transient and then analyzing with FFT the spectral content outside the band we can investigate intermodulation distortion created in a transient way!!! By playing with f1, f2, T and the amplitude of the pulse then analyzing the distortion I believe a powerfull test can be designed.
Let me know what you think.
JPV
I thought you guys might enjoy my latest LME49811 with Allison class A/B output - the device stays in class A to about 10watts or so. (It's actually adjustable). My LME49811 driver board measures at .0035 Thd. my V6a allison measures .0033. It's driven with the sink/source of the LME, the version V7 combines the sink source and drives between the allison emitters and uses bootstrap current supply. More info can be found in the Simulation of Unique Allison thread in the Solid State Forum.
These are lower power measurements. I'm waiting for a new sound card to preform the high power tests.
Ken L
These are lower power measurements. I'm waiting for a new sound card to preform the high power tests.
Ken L
Attachments
I do not see the advantage of using a Vbe multiplier with TT. The Vbe multiplier is essentially used to take advantage of the Vbe tracking because it is a transistor like the output ones. In TT's the sensor are the diodes and the Vbe transistor is not used as a sensor so what's the purpose certainly if the diodes are fed by a stable current source like the lm49810.
Yes I have used this pulse for other measurements in my Master thesis at the university long time ago. Tooling was heavy ( PDP11 for real time and IBM 370 for FFT calculation !!!, software to be done by yourself )
Later I had the opportunity to try ( on loan from a friend with HP) the HP first portable FFT analyzer and I made some experiments using this pulse in the way explained to generate and analyze distortion created in a transient way and this for amplifiers and speakers. We used BGW 100 amplifier and Altec Lansing model15. As I recall, we saw distortion components which were different than the steady state figures but I had to give back the instrumentation and didn't pursue. Then jobs in different direction...
I always wanted to go back to these tests and now that cheap instrumentation is available, it can be reinvestigated home !!
It is on my to do list.
JPV
AP was not existing at that time ( begin 70thies).
I have read the papers from AP on multitone and it is of course a great tool.
I am surprised not to see more tests results with this, did you use it?
What I will try is more a transient generator with controlled bandwith which is in a way antinomic because transients have by essence large bandwith.
Controlling the bandwith, the location of the bandwith and the amplitude of the pulse allows you to trigger nonlinearities in an interesting way.
As for the Vbe multiplier, I have tried to use its multiplication factor but could not find practical biasing to allow the tuning. The multiplication factor multiplies everything and we need separate control of a static bias voltage source and a drifting voltage source.
The only 'interesting use' I see is with the 49811 wher you can use a Vbe multiplier with a resistor in the collector ( Hawksford). This has the advantage to make the Vbias less sensitive to collector current and voltage variations in the VAS transistor.
With the LME49810, this topology is impossible to implement but the stable current source is making this not necessary I guess.
JPV
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.