T-network: the better feedback solution?

Status
Not open for further replies.
Jump to Latest
#322
Sheldon,

WRT the buffers' supply, you may also try to load additionally its output (each rail to ground) to get somewhat lower output impedance. Assuming one supply for both buffers and 5mA per buffer, you may use something about 2k. Lower values may be better, those below 500R (for 470R 1W would be recommended minimum) will bring the output impedance below 1 Ohm, but then some heatsinks on the second transistors would be needed.

Pedja
 
#324
Nuuk said:
Good work Sheldon!

You may care to try the single cap LPF as both Pedja and I prefer it to the two cap version!

I have built two TNIGC's and they are both dead silent. If I stick my ear to the (high efficiency) speaker cones all I can hear is the hum from Carlos FM's amp in Portugual! :bigeyes: (OK - a little exaggeration there but you know what I mean 😉 )

Did you hand pick your T-network resistor values? I went with the 10K-100R-10K and then adjusted the NI to ground value. I got under 3 mV on all four channels (NB two amps).

I am yet to try my regulated supplies with the discrete buffers and T-network. (Too :smash: )
Click to expand...


The tiny residual hum may be due to the very compact nature of the whole amp. I suppose I could shield the AC lines, but I'm happy with less than 1 ma of AC output. Also, for some reason, I have about 10v DC on my house lines here. I know that can sometimes cause mechanical transformer hum, but I don't think it would increase AC field noise?

I just picked network values from what I had on hand and stuck them in there. I'm not inclined to mess around in order to lower the offsets. Too many other things to try, like the alternate LPF values. But I might try about 108R in the T to bring the gain a little higher, as Franz suggests.

Sheldon
 
#325
Pedja said:

That is the good idea for the supply, just don’t forget to remove the resistors from the emitters of the second BD139/140.


To clarify my opinion... I do like 75R-47nF better from the tonal point of view, but the shown second order filter had better soundstage.

Pedja
Click to expand...

Yes, no resistors on emitter of the second transistor in the pair.

Sheldon
 
#326
Pedja said:
Sheldon,

WRT the buffers' supply, you may also try to load additionally its output (each rail to ground) to get somewhat lower output impedance. Assuming one supply for both buffers and 5mA per buffer, you may use something about 2k. Lower values may be better, those below 500R (for 470R 1W would be recommended minimum) will bring the output impedance below 1 Ohm, but then some heatsinks on the second transistors would be needed.

Pedja
Click to expand...

Thanks, I had thought about that. Based on my rudimentary knowledge, it seemed from intuition that a little more load on the supply would put the output transistors in a better part of their range. But I wasn't sure if it would make a difference. Given your encouragement, I'll try it with the higher values you suggested. My transformer may run out of juice if I go too low on the resistor. Plus, I didn't leave much room for heatsinks.

Sheldon
 
#333
Back on topic, then

I'm testing with an LM1875 module I made some time ago, tested and :hot:ed-in, but still unfinished (no case).
I removed the feedback resistors and kept the 10k input resistor, so I've chosen different values for feedback.
These are the results:

LM1875 amp:

Original:
Rg: 10k
Rf: 220k
NI direct to ground

Left channel: 13mv
Right channel: 3.7mv

----------------------------------------------------

T-network:
R1: 10k
R2: 4.7k
R3(to ground): 50
R4: 4.7k

With NI direct to signal ground:
Left channel: 95mv
Right channel: 71mv

With NI through 50k MT-pots to signal ground:
Left channel (set to 20k): 0.0~0.1mv
Right channel (set to 14.8k): 0.0~0.1mv

When I replace the MT-pots by resistors I will choose smaller values, even if I will have some DC-offset (10~20mv will be fine).

I will finish this maby tomorrow night and listen in a few days.
 
#334
Carlos

It would be interesting, with the LM1875, when you could compare t-net and conventional feedback with this chip.

I am not sure, if the difference is as big as with the "Ouverture" series of LM chips.

Why?

I changed my mind about the effect of the t-net. I am now with this explanation:

phase_accurate (eg Charles) wrote: Where it differs is the impedance in question. For the same gain a lower impedance feedback path can be used with the voltage divider network. This leads to reduced susceptibility to parasitic capacitances (including the amps's input node which is not even a linear capacitor !) and therefore better accuracy of the signal fed back to the input, including errors caused by the load.
Click to expand...

Remark the part from the parasitic capacitances!

BTW: Charles: where in Switzerland do you live? Maybe, we could meet one day?

Franz
 
#337
There are other advantages in using the t-network topology.
Smaller resistors also mean less noise and distortion.

Anyway, I can't find anything about parasitic capacitances on the datasheets.
The datasheets don't even spec the input capacitance of these chips...
Can you elaborate?
Do you have any specific information about this for the LM3875 that's not on the datasheet?
Or did I miss some posts here?😕
 
#340
Carlos

Maybe, I missed it: did someone try the t-net with LM1875 before?

If not, thats why I cannot answer your question. And it would be interesting, if the difference is as big as with other LM chips.

It should be easy for you, to change the circuit from t-net to conventional feedback,to compare (for us). I am wrong?

Franz

P.S.
I want to show you my workbench tonight:

An externally hosted image should be here but it was not working when we last tested it.
 
Status
Not open for further replies.
Top Bottom