Instability on a diferential output amplifier

[Miguel Lopez]

Hello

I've been trying a diferential output amplifier, as it is my intention in the future to build a transistor oscilloscope, with better performance that the valve one that I'm currently using. Althought the circuit that I've been testing amplifies the input signal very well, it suffers from instability with no signal applied.

That means that with the input connected to ground, the voltage across points A and B (see picture) is constantly "moving". The range in which it moves, is about 0,6V. Then the trace on the scope will be moving too.

Voltage at points C and D, are very stable, and Vcc (60V) is also very stable. So I can not find tha reason for this instability.

I'm asking for help to solve such instability.

Thanks in advance

 

[BobK]

I don't have an answer, so I will ask some questions.

Why are the collector resistors for the differential pair connected to +60 instead of the ~55V regulated rail?

Are the differential pair matched and kept at the same temperature?

The 1K collector resistors seem like they are too low. You have a current source of approximately 6mA. When the two inputs are at the same voltage (normal operation of a differential amp) they will share this current, so 3mA each. Thus they drop 3V from the 60V supply. I would think you want them to drop about half the supply, so 10K would be about right.

Bob

 

[shrtrnd]

My thoughts are around BobK's observation that it'd be nice to have a matched pair on the transistors, and keeping them at the same temperature is critical. They make dual heat-sinks for this purpose.

 

[BobK]

Looking at it again. Here is my analysis of the differential pair's operating point:

The voltage at the emitters is about 4.7 - 1 diode drop or 4V.

The voltage at the collectors is about 60 - 3 = 57V.

The current through the transistors is ~6mA. So we have them dissipating:

P = I V = 0.006 * 53 = 0.318W

This is enough to significantly heat them, and if they are not thermally coupled that would probably explain the drift. In fact is is probably close to the max dissipation for a TO92 if that is what you have.

I think you want a much lower design current.

Bob

 

[Miguel Lopez]

Have to apologize. As BobK noticed and explain in the post #2, the collector resistor of 1K are very low. Indeed, that is a mistake while drawing the schematic. They are in fact 10K, as BobK suggested. Sorry for that.

Why are the collector resistors for the differential pair connected to +60 instead of the ~55V regulated rail?

Sorry, I do not understand this question. Which 55V rail?

Are the differential pair matched and kept at the same temperature?

Matched, no. At the same temperature, well, I situated them one besides the other, for this, but they are not mounted on any heatsink. The transistor are Soviet 2T602 retrieved from old Soviet oscilloscopes where they did the same function as in this circuit.

I will try a heatsink for both pf them.

Thanks

 

[Miguel Lopez]

Here the picture updated.

 

[Arouse1973]

I have never seen a diff amp biased with a zener current sink before, why the choice?
Adam

 

[Miguel Lopez]

I have never seen a diff amp biased with a zener current sink before, why the choice?
Adam

Well, while in the College, the professors always said that it is better to bias a long tail pair using a constant current source, as the resistance seen by the diferential pair from their emitters to ground is almost infinite. I can find the proper explanation in my books, but I don't remember why right now.

I could use an LM317-based current source, but I must use what I have at hand.

 

[Arouse1973]

Yes you are correct, and you would normally use a mirror circuit and not a zener. Unless it was 5.1V zener which has zero temp-co.
Adam

 

[shrtrnd]

Miguel.
Just heat sinking them is not what is needed in this case. They need to be thermally sinked TO EACH OTHER.
The heat sink between the two is usually accomplished with a sink that JUST connects the cases of the transistors together. They make sinks specially for this purpose. The transistors need to be both at the same temperature.

 

[BobK]

Sorry, I do not understand this question. Which 55V rail?

Sorry, I substituted the 4.7V zener for the 24V Zener in my head. I now see that is rail is 24V.

But what I said about the dissapation still applies. I think if there is any heating of the differential pair you are going to get drift problems. Is there any reason you cannot lower the current through them?


Bob

 

[Externet]

Welcome back from 'excomunication' !
Not to answer you question about stability; but to ask you to look a little beyond the road on your project.
Perhaps you will come across an orphan LCD monitor and will change idea on using such as display, perhaps all the praiseworthy effort and time to be spent in your attempt now will change direction.
What size screen would you be pleased with ? What bandwidth would you be pleased with ? Will you want at a certain point storage features ? How many channels would make you happy ?

This is ridiculously priced :

----> http://www.aliexpress.com/item/DSO1..._1&btsid=43e2b524-0245-4428-bc86-24e73e97926c

If you cannot get it there, I can send/forward it to you from USA...

Others :
----> http://www.ebay.com/itm/DSO-Nano-v3...677960?hash=item51e560cf48:g:Q84AAOSwNsdXQlAn

----> http://www.ebay.com/itm/GW-Instek-E...359568?hash=item2c9e4c4510:g:zcQAAOSwxN5WWBi7

----> http://www.ebay.com/itm/LCD-Digital...149149?hash=item1ea4dfcc5d:g:JfMAAOSw37tWCfrz

And do not forget that these humble animals exist :
----> http://www.dxzone.com/catalog/Software/Oscilloscope/
----> http://www.zelscope.com/
----> http://www.zen22142.zen.co.uk/Prac/winscope.htm
! This is a downloading link !----> http://www.sillanumsoft.org/Download/VAsetup.exe

And if any member of the forum has a spare osciloscope to donate, I will pay shipping to Miguel. I do not have a spare now.

 

[Miguel Lopez]

Unless it was 5.1V zener which has zero temp-co.

I will do that. I have some 5.1V Zeners. But I think that in that case I would need to add a forward biased diode in series with the Zener, to compensate the thermal drift of the base-emitter junction of the transistor, wouldn't I?

But what I said about the dissapation still applies. I think if there is any heating of the differential pair you are going to get drift problems. Is there any reason you cannot lower the current through them?

I can make the current lower, no problem on that. Let's say, 1 mA on each transistor at balance condition. That is 2 mA in the constant current sink.

They make sinks specially for this purpose. The transistors need to be both at the same temperature.

this is a bit harder, as I searched in my stuff, and I have no heatsink for two transistors. I can make one but there is the problem that they has to be electrically isolated. I will see what I can do. On Soviet scopes they are not attached to the same heatsink, and there is no drift. Of course, those Soviet scopes are very much complex than this circuit.

What size screen would you be pleased with ? What bandwidth would you be pleased with ? Will you want at a certain point storage features ? How many channels would make you happy ?

I will use a CRT similar to the one that I used for the valve scope. Bandwidth, up to 50 kHz will be very well for me. Storage, NO!!!. Channels, only one. Ijust need it to go on with my experimentation, and I usually never goes out of the audio spectrum.

And if any member of the forum has a spare osciloscope to donate, I will pay shipping to Miguel. I do not have a spare now.

Thank you very much for your kind offer Externet, but there is a little problem. There is the Cuban Customs in the middle. A scope will be taxed as a equipment and I would have to pay a BIG sum of money (at least for my pocket), to extract it from the post office, so I prefer you don't do that. Thanks anyway.