PRR

Re: Small Signal Diode Differences
« Reply #20 on: June 09, 2020, 01:26:51 AM »
> offset is largely determined by the bias current in the LTP. Grrrrr.

By the current-match across the pair.

If you take only one output from the pair, yes, you can "trim" the tail current. Like if your boat is bent, you may be able to ballast it to go straight. But it may be more elegant (and lower THD) to make both sides the same.


apzx

Re: Small Signal Diode Differences
« Reply #21 on: June 09, 2020, 03:56:53 AM »
> offset is largely determined by the bias current in the LTP. Grrrrr.

By the current-match across the pair.

If you take only one output from the pair, yes, you can "trim" the tail current. Like if your boat is bent, you may be able to ballast it to go straight. But it may be more elegant (and lower THD) to make both sides the same.

Yeah, the problem with the M100 design is that the "tail" current (that is the current flowing from the collectors of the input pair, yes?) for the non-inverting input goes through a 4k7 resistor to the negative rail and the inverting input goes straight to the negative rail. So, as the LTP is run richer with more current the balance between the two collector currents gets more and more imbalanced. I did not notice this at first.

I looked at the JE-918 and that uses a current sink on the non-inverting input and an 8k2 on the inverting input. Granted the transistors are matched for Hfe and Vbe, and it is quoted to have an offset of <5mV. However, for playing around I decided to go a step further. So, on the M100 the current for the LTP and the BC177B on the right that looks like it sets the operating current of the output stage are all set via a 22k resistor. I replaced it with an active current sink. Then for the LTP I added a current mirror, and that really forced the current differences to bugger all. Here is my current schematic. The LTP is running around 960uA with the tail current differences being some 200pA. The output stage re-looking at it is running a bit richer than I'd like at about 3.8mA. I need to configure the Vbe multiplier to drop the current down to about 3mA, just haven't bothered to yet. This is a rather fun exercise though, and I have learned A LOT about how these little devices work on the inside, and it just gives me more appreciation for the work that goes into them. Really, very enlightening stuff!

JohnRoberts

Re: Small Signal Diode Differences
« Reply #22 on: June 09, 2020, 09:24:11 AM »
Yeah, the problem with the M100 design is that the "tail" current (that is the current flowing from the collectors of the input pair, yes?) for the non-inverting input goes through a 4k7 resistor to the negative rail and the inverting input goes straight to the negative rail. So, as the LTP is run richer with more current the balance between the two collector currents gets more and more imbalanced. I did not notice this at first.
It would be far easier to trim out offsets. 

The 4.7k and Vbe roughly established the LTP balance for 0V output when base drive of the common emitter voltage gain stage equals the class A current of the pre-output stage. The lower value 680 ohm vs 2.2k local feedback R will affect equilibrium and open loop gain. 
Quote
I looked at the JE-918 and that uses a current sink on the non-inverting input and an 8k2 on the inverting input. Granted the transistors are matched for Hfe and Vbe, and it is quoted to have an offset of <5mV. However, for playing around I decided to go a step further. So, on the M100 the current for the LTP and the BC177B on the right that looks like it sets the operating current of the output stage are all set via a 22k resistor. I replaced it with an active current sink. Then for the LTP I added a current mirror, and that really forced the current differences to bugger all. Here is my current schematic. The LTP is running around 960uA with the tail current differences being some 200pA. The output stage re-looking at it is running a bit richer than I'd like at about 3.8mA. I need to configure the Vbe multiplier to drop the current down to about 3mA, just haven't bothered to yet. This is a rather fun exercise though, and I have learned A LOT about how these little devices work on the inside, and it just gives me more appreciation for the work that goes into them. Really, very enlightening stuff!
There are other topology variants that balance easier, but is the complexity worth it?  You may want to look at audio power amp designs that mainly differ in output stage and PS. Should be more examples to look at. 

JR
Don't only half-ass tune your drums. Visit https://circularscience.com to hear what properly "cleared" drums sound like.

PRR

Re: Small Signal Diode Differences
« Reply #23 on: June 09, 2020, 06:15:31 PM »
> goes through a 4k7 resistor to the negative rail

In that configuration, counting on thumbs, the "4.7" should really be twice the 2.2k in the tail source. Since 2*2.2 is very nearly 4.7, this should come out very close and avoid driving over to the store to find 4.4k or 2.35k or whatever.

When you get that far you find that thumb-math is imprecise. The Jensen 990 got that far and then found he had to add diode(s) to force Vce equal because Early Effect left a residual mV offset on a zero-offset pair. (And it mattered because he was DC-coupling soft iron output transformer.)

Your mirror not only FORCES current match, it doubles the gain.

Douglas Self and Bob Cordell devote whole chapters to input pair topology and value-picking.

apzx

Re: Small Signal Diode Differences
« Reply #24 on: June 10, 2020, 06:16:01 PM »
JR & PRR you two are absolute fountains of knowledge!

So, I have not really made any major advances in things because I am still trying things, but I have been playing around with different things in the scamulator. I had not considered the increase in the open loop gain in general by changing these things out. I think the M100 is supposed to have about 90dB. My current iteration sitting in the simulator has like 126dB, but I am not "necessarily" trying to improve the open loop gain right now. What I am trying to do now is understand how to adjust to the LTP tail current and more or less balance it.

I played around with the 2k2 current source resistor and the 2k2 "local feedback resistor". On the M100 the VAS and PNP driver is one area where I do not fully understand what is going on. It basically looks like the BC177B is mostly acting as a driver/buffer for 2N2905 output. Regardless, I played with some values just to see what happened, and I found it rather interesting that the offset can sort of be adjusted just by varying those two values. Just for the sake of experimentation I took the JE-990's tail current design, inverted it, and played around with it just to see if I could get the currents to more closely match one another. And yeah I got it to much more closely match on the tail currents. Though unlike the JE-990 the resistors have to be different values, but the best case I got was when the total value of the resistors was roughly 2x times the value of the 820R resistor on the current source (1.66k is the total value I ended up with). Additionally, the difference between the two resistors happens to be right around the difference between the 680R in the VAS/buffer and the 820R LTP current source. The non-inverting input got a 750R and the inverting input got a 910R plus a 1N4148. I am ASSuming that there is a reason for this that likely relates to the ratio between the VAS/buffer stage and LTP current source imbalance that I ended up with. Unless it is just a coincidence that things like this just happen to match up which I find unlikely.

I've been looking at some power amp stuff as JR suggested and just seeing how they do things a bit differently. I've got a couple of different ideas that I want to try, but just have not quite gotten around to it. Again seriously thank you for taking the time to write responses. I greatly appreciate the assistance on this. I have gained so much appreciation for opamps in general by tinkering with this circuit.

Re: Small Signal Diode Differences
« Reply #25 on: June 10, 2020, 09:38:06 PM »
Just use a current source for the LTP emmiters and a current mirror on the collectors and move on. Using a current mirror not only has the advantage of balancing the collector currents which in turns lower distortion, but also, the transconductance of the LTP is doubled, the current mirror is acting as a differential to single ended coverter.

Like PRR mentioned, check out power amplifier books, with the exception of the output stage the rest is exactly the same to an opamp, both Douglas Self and Bob Cordell explain things very well in their books. Also check out books on integrated circuit design, the one by Sedra/Smith and the one by Gray/Hurst/Lewis/Meyer are excellent and comprehensive.

Also, if you haven't read it yet, you should, Nelson Pass has a good introductory article on DIY opamps, check it out https://www.passdiy.com/project/amplifiers/diy-op-amps


 

Related Topics

  Subject / Started by Replies Last post
18 Replies
19260 Views
Last post March 02, 2006, 07:52:40 AM
by AMZ-FX
7 Replies
2280 Views
Last post March 31, 2006, 03:33:53 PM
by Matt Syson
16 Replies
23162 Views
Last post December 18, 2010, 10:57:46 AM
by electrochronic
16 Replies
2330 Views
Last post January 15, 2011, 10:40:19 PM
by dmlandrum