Maybe the fb network sucks major part of output current
Maybe not. Honestly here I don't get it right with the design, if the author could give us some lights, please
Regards,
nAr
I am back from traveling and just began to get caught up on the thread questions that came up while away.
I'm glad to see that there is interest and circuits being built!
A first order LPF does exist at the input circuit and it sounds like you are measuring the response. The input capacitor forms a first order rolloff that is dictated by the feedback network impedance and the capacitor value. With 1uF input caps the -3db point is roughly 18Hz and the circuit is -0.1db at about 100Hz. Double them to 2uF and the -3db point drops to around 9Hz. If we use 10uF caps at the input like the aleph P had, the roll off at a gain of 5 looks to be -0.1dB somewhere around 10Hz and the -3dB point is around 2Hz. With 10uF caps on both inputs, frequency response should be pretty flat way down low.
It’s possible there is a measurement setup contributor to the rolloff if it does not follow the expected behavior above. What is the output impedance of the generator?
Try connecting the scope at the input before the capacitor and make sure that the sweep is flat through the frequency range of interest. Make sure 50ohm termination is off on the scope if you have not checked. It is also possible that the generator is not flat when loaded, although the above test should also eliminate that as a concern. Just suggesting some things to build confidence in the measurements.
Not sure I answered the question but I can’t see any other problems with it so far. My test case hardware is in progress but waiting on a few parts yet. I will catch up as quickly as I can. Once there I will post some graphs.
Dave
I'm glad to see that there is interest and circuits being built!
Hi Bill,
A first order LPF does exist at the input circuit and it sounds like you are measuring the response. The input capacitor forms a first order rolloff that is dictated by the feedback network impedance and the capacitor value. With 1uF input caps the -3db point is roughly 18Hz and the circuit is -0.1db at about 100Hz. Double them to 2uF and the -3db point drops to around 9Hz. If we use 10uF caps at the input like the aleph P had, the roll off at a gain of 5 looks to be -0.1dB somewhere around 10Hz and the -3dB point is around 2Hz. With 10uF caps on both inputs, frequency response should be pretty flat way down low.
It’s possible there is a measurement setup contributor to the rolloff if it does not follow the expected behavior above. What is the output impedance of the generator?
Try connecting the scope at the input before the capacitor and make sure that the sweep is flat through the frequency range of interest. Make sure 50ohm termination is off on the scope if you have not checked. It is also possible that the generator is not flat when loaded, although the above test should also eliminate that as a concern. Just suggesting some things to build confidence in the measurements.
Not sure I answered the question but I can’t see any other problems with it so far. My test case hardware is in progress but waiting on a few parts yet. I will catch up as quickly as I can. Once there I will post some graphs.
Dave
Hi nAr,
The circuit has some complexities I don't yet fully understand either but am anxious to discover. I tried a number of different current sinks, cascoding of the sinks, cascodes on the pair, current sources replacing the resistors, and even an additional stage while modeling. I hoped to keep it simple and found that with only NJFETs in this configuration, everything kind of fell into place.
When modeling different current sinks for example, the predicted performance suffered considerably even though the current sinks themselves were "better". Best I can figure is that the gain devices prefer the current sinks to behave similarly, not pushing or pulling any harder than themselves for a given current modulation. If it turns out to be true, kind of makes sense I think.
Until we can listen to variations of it, I can't predict what it will sound like or suggest an improvement... If Bill does not find any big holes in the circuit, perhaps I will do a quick PCB we can do some listening tests with.
Dave
The circuit has some complexities I don't yet fully understand either but am anxious to discover. I tried a number of different current sinks, cascoding of the sinks, cascodes on the pair, current sources replacing the resistors, and even an additional stage while modeling. I hoped to keep it simple and found that with only NJFETs in this configuration, everything kind of fell into place.
When modeling different current sinks for example, the predicted performance suffered considerably even though the current sinks themselves were "better". Best I can figure is that the gain devices prefer the current sinks to behave similarly, not pushing or pulling any harder than themselves for a given current modulation. If it turns out to be true, kind of makes sense I think.
Until we can listen to variations of it, I can't predict what it will sound like or suggest an improvement... If Bill does not find any big holes in the circuit, perhaps I will do a quick PCB we can do some listening tests with.
Dave
After consulting with myself, I realized I am stupid. I hadn't changed both input caps to 10uf.
This is what I have using a dual trace Tek, 600ohm gen @1Vrms in single ended with minus input to ground.
At 30hz barely perceptible decrease in output.
At 25hz output is down around 100mv peak.
Plus and minus outputs agree.
I'm thinking I could scale the feedback network up, any thoughts?
Should I also change R9?
This is what I have using a dual trace Tek, 600ohm gen @1Vrms in single ended with minus input to ground.
At 30hz barely perceptible decrease in output.
At 25hz output is down around 100mv peak.
Plus and minus outputs agree.
I'm thinking I could scale the feedback network up, any thoughts?
Should I also change R9?
Last edited:
I tried transformer coupling the gen to the inputs with a Jensen line trafo, omitting the input caps.
The circuit is flat but loads the trafos down at low freq, the input impedance must be around 1k. This ain't gonna work for me, I use trafo OP dacs. Looks like you would need a source with less than 100ohms output impedance to work, but I may be wrong.
Any ideas out there?
The circuit is flat but loads the trafos down at low freq, the input impedance must be around 1k. This ain't gonna work for me, I use trafo OP dacs. Looks like you would need a source with less than 100ohms output impedance to work, but I may be wrong.
Any ideas out there?
buzzforb,
Sounds like a great combo. I can’t wait to hear how it sounds.
Bill,
Test setup looks fine. You will probably notice that one output will have slightly less signal than the other when inputting an SE source. Changing R9 effects this but also increases distortion as the value decreases. I balanced the tradeoff between the two. I definitely optimized R9 for the different gains. Increasing the input network values up is ok to a point but begins to fall on its face the further you go. The original values for different gains are reasonably well optomised if you can live with them.
Bill,
I did a little experimentation with source impedances up to 1K ohm. Aside from the voltage divider formed by the generators source impedance and the feedback network causing the input level to be lower across the entire band, I did not see much of a shift in the frequency response.
I wonder if the LF loading is caused by the input current to the transformer or a primary side mismatch instead of the outputs connected to the line stage? The input impedance of the stage is pretty straight forward and flat. Rolloffs are dictated by the input/output caps and on the high side by C1/C2. The input current at gain of 5 to the line stage looks like ~65uA per side for 1VRMS out differential. I really have no transformer experience in this context so take it for what it is worth…
Dave
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.