To lift a bit Vbe multiplier voltage, otherwise thermal compensation with TT diodes is undercompensated. I simulatid it with normal diode and schottky diode was better suited with its lower voltage drop.
dado
Can I rephrase that reply?
Without the added didoe the Vbe output is at the required bias voltage. That also multiplies the diode compensation by the same ratio.
With the added diode D8, the Vbe output voltage is the same but the multiplier is at a slightly lower ratio. This results in the diode compensation being lower due to the lower multiplier ratio.
The added diode reduces the temperature compensation. Without the added diode, the bias voltage would be over compensated.
That last conclusion is opposite to what you said.
Without the added didoe the Vbe output is at the required bias voltage. That also multiplies the diode compensation by the same ratio.
With the added diode D8, the Vbe output voltage is the same but the multiplier is at a slightly lower ratio. This results in the diode compensation being lower due to the lower multiplier ratio.
The added diode reduces the temperature compensation. Without the added diode, the bias voltage would be over compensated.
That last conclusion is opposite to what you said.
I think you will find that you can "tune" the temperature compensation, without incurring the "step change" of using discrete added diode/s.
Simply change the current in the TT diodes and the compensation changes. This is easily done by adding a VR parallel to the diode string. This VR bypasses some of the current around the diodes.
Simply change the current in the TT diodes and the compensation changes. This is easily done by adding a VR parallel to the diode string. This VR bypasses some of the current around the diodes.
No is the answer on both!
Thermal compensation has two part, TT diodes to compensate thermal drift of the output transistors and Vbe multiplier to compensate thermal drift of the drivers(that is why Vbe multiplier is on the same separated small heat sink). To be effective the Vbe multiplier needs more voltage and this is why D8 was added. I this case we have four transistors Vbes plus D8 diode voltage drop, and that is close to 3.2 V(not 2.8 without diode).
Without that diode drivers are undercompensated, with diode drivers are a bit overcompesated(better then temperature run away).
I have a resitor parallel to the TT diodes for fine tuning.
I simulated this thermal compensation using Cordell models for output, tt diodes, drivers and Vbe transistors, but I need to test it in real amp, not sure when as I am occupied with other project. Maybe somone can do that?? I tried in my simulation Cordell's Vbe multiplier(figure 14.19a page 307) but it a was way overcompensated according to the simulation. Could be that the simulation(models) is not acurate enough, so practical test will be neaded.
Cordell, in his book describes Vbe multiplier for output transistors only, drivers not being mounted on the main heatsink, neither Vbe transistor. Hi mentioned that if predrivers are mounted on the main heatsink then Vbe transistor should be there too.
I want predrivers on separate heatsink and compesated with Vbe transistor on the same heatsink.
dado
Thermal compensation has two part, TT diodes to compensate thermal drift of the output transistors and Vbe multiplier to compensate thermal drift of the drivers(that is why Vbe multiplier is on the same separated small heat sink). To be effective the Vbe multiplier needs more voltage and this is why D8 was added. I this case we have four transistors Vbes plus D8 diode voltage drop, and that is close to 3.2 V(not 2.8 without diode).
Without that diode drivers are undercompensated, with diode drivers are a bit overcompesated(better then temperature run away).
I have a resitor parallel to the TT diodes for fine tuning.
I simulated this thermal compensation using Cordell models for output, tt diodes, drivers and Vbe transistors, but I need to test it in real amp, not sure when as I am occupied with other project. Maybe somone can do that?? I tried in my simulation Cordell's Vbe multiplier(figure 14.19a page 307) but it a was way overcompensated according to the simulation. Could be that the simulation(models) is not acurate enough, so practical test will be neaded.
Cordell, in his book describes Vbe multiplier for output transistors only, drivers not being mounted on the main heatsink, neither Vbe transistor. Hi mentioned that if predrivers are mounted on the main heatsink then Vbe transistor should be there too.
I want predrivers on separate heatsink and compesated with Vbe transistor on the same heatsink.
dado
For this amp only simulation, but something similar I used in my TT TRIPLE amp and that amp was built.
As English is not my native language I am not good in explanation. Do you have Bob Cordell book? Its good reading.
A bit late
Hi Damir
Sorry this is a bit late but I have had a few ideas lately that may be relevant.
Have not simulated to see if the theoretical improvements make any practical difference but may do so in a few days.
.
If you increased the Current Mirror emitter resistors then you would have less noise and also it may then be possible to balance the mirror better.
I don't think you need TO220 drivers. TO126 devices like 2SC3503 have adequate power capacity (SOA) and are faster so a little more feedback could be safely applied.
Best wishes
David
Hi Damir
Sorry this is a bit late but I have had a few ideas lately that may be relevant.
Have not simulated to see if the theoretical improvements make any practical difference but may do so in a few days.
.
If you increased the Current Mirror emitter resistors then you would have less noise and also it may then be possible to balance the mirror better.
I don't think you need TO220 drivers. TO126 devices like 2SC3503 have adequate power capacity (SOA) and are faster so a little more feedback could be safely applied.
Best wishes
David
Hi David,
Good ideas, I will try it, at least in simulation first. I know about CM emitter resistors, but I did not think about it when simulated Little GEM:
BR Damir
Isn't (K)SC3503 a tad lean on current rating (100mA) for a driver?
As a video driver and VAS transistor, sure.
Maybe 2SC4793 if you prefer light and fast drivers?
http://www.datasheetcatalog.org/datasheet/fairchild/KSC3503.pdf
http://www.datasheetcatalog.org/datasheet/toshiba/3220.pdf
Hi David,
Have you seen my pure Cherry simulation with Little GEM where I use 330 ohm for CM emitter resistors. http://www.diyaudio.com/forums/solid-state/235188-tpc-vs-tmc-vs-pure-cherry-17.html#post3497689
BR Damir
Why not?
I have not done any SOA checks but Dado's amp only uses about 50mA so it looks OK at first impression.
He already had one 2SC3503 so that was just the first possibility.
The point was that the optimal driver transistor does not have to be a typical TO220 just because it is labelled a "driver"
Or rule out a transistor just because of a "video" label.
Best wishes
David
Last edited:
Hi Damir
I saw that but I concentrated on the different compensation and did not notice the altered CM resistor values.
It inspires confidence if someone else comes to the same conclusion
The altered resistance can affect the CM balance.
I think R14 should be lower but I can't find the BC546 model?
Best wishes
David
Hi Damir
I saw that but I concentrated on the different compensation and did not notice the altered CM resistor values.
It inspires confidence if someone else comes to the same conclusion
The altered resistance can affect the CM balance.
I think R14 should be lower but I can't find the BC546 model?
Best wishes
David
Hi David,
In that simulation I used BC546B from LTspice library. Maybe better to use BC550C of the Cordell models.
BR Damir
Thanks.
The mirror is best balanced when the base current of the VAS EF is equal to the total base current of the two CM transistors.
With the beta of the BC546B about half of the BC550C that means the emitter resistor of the BC456B should be about the same as for the CM.
The extra current should also mean it is a bit faster so less contribution to phase.
Best wishes
David.
Why?
This amp is specified 100W into 8 ohm, not rated into lower impedance, so nominally 5 amps peak as a quick approximation.
The output transistors maintain typical beta around 100* so the drivers need around 50mA peak. So 100mA continuous should leave some headroom.
A proper SPICE simulation and conservative belief may show the need for a bit more but it doesn't appear to be too far from the capacity of the 2SC3503.
If more is needed then the 2SC3601 is rated 150mA continuous.
Best of all the 2SC4623 at 300 mA continuous (and 400Mhz Ft).
Isn't that what I just wrote?
Best wishes
David
*I don't use the OnSemi outputs if they are less than 100 hFE. Easy to select when only one pair.
Last edited:
the maximum output current is not 5Apk for a 100W amplifier feeding an 8ohms speaker.
It is more likely to be at least 10Apk and can for very fast transients approach 25Apk. I design for 15Apk.
Look at the output device gain when supplying a 15Apk current. It is more likely to be around 10times to 50times. If you use multiple pairs you get a BIG increase in the effective hFE. It is here that Self's Invariant shows up the better distortion performance of the multiple pairs feeding the 4r0 load rather than the designed for 8r0 load.
15Apk with a 10 hFE requires a 1.5Apk from the driver. But rarely do we require the driver to actually pass 1.5Apk
The driver needs investigation of the temperature de-rated SOAR into reactive loading just like the output devices. It's from this that you find that even a 1A to 1.5A driver needs lots of dissipation capability to be able to drive that reactive load. A 150mA driver cannot survive transient currents into reactive loads.
Yes, but I went on the clarify the effect to look for when the CM is doing it's job.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.