Rush Cascode Headphone Amp + JLH Output Stage

Status
Not open for further replies.
Hey keantoken,

Maybe you could arrange another type of offset compensation. To be useful I don´t think your very nicelooking circuit should be so dependant of the ohmic values of the feedback network.

No, there´s nothing wrong with my file. Just removed all your variables to make the dc/ac/tran-sims easier and more straightforward. Will play around some more when I get home from work and also make a LTP comparision to the NTP(perfect name😉). That one I will post. Actually I have done the LTP both with BJT´s and FET´s before. Just a matter of copy and paste and adjusting to your sand-models, currents and voltages.

Great fun anyway.
 
Hey keantoken,

Interesting circuit! Do you have any schematics of NADs application?

Did some DC-sims with the circuit in your first post but can not eliminate some DC at the output(in the ballpark of 60mV). The optional DC-comp didn´t work as supposed. Tried both with DC and AC-feedback.

Anyway THD with 20db gain seems very good. But that is probably what one could expect with an open loop gain of ca 66dB.

Will do some sims using a PNP LTP for comparision.

Hello

If you mean the Nad phono preamp using a rush (ntp) input and designed by Tomlinson Holman.

Here it is for you and keantoken to save it on yours hard disks.

Bye

Gaetan
 

Attachments

  • nad3020pre.gif
    nad3020pre.gif
    87.9 KB · Views: 593
Last edited:
You're right Forr, thanks. I was on the verge there but just hadn't realized it was this simple. The two are directly related, only opposites that perform the same function.
This gives rise to some more naming possibilities.
- keantoken

In this forum, someone stated that all single amplifying elements (tubes, bipolar or FET transistors) should even be considered as differential devices, reacting to a voltage between two inputs. This way to see circtuits greatly simplifies their analysis.
 
The base current issue may make it seem "dirty", but when including input/output caps, output is still very stable with temperature. Increase R2 and R3 to 10k and offset is easily adjusted again.

Could still be problematic though...

- keantoken
 
Also, how do I get around the hum problem? I could give more detailed data if I could connect it to my computer with minimal noise. Any ideas?
- keantoken

It's probably your grounding. If I were you I'd use a couple of 9V batteries in series, with the middle point being 0V, and connect the source ground to this 0V point. Keep all probes off for this test, including DMMs that are not battery operated. If you got rid of the hum thus the next step is obvious.
 
Just so's you know: JLH Class A is slightly (or outrageously depending the
output transistor) more wasteful of quiescent current than "normal" class A.

Class A output current would have crossed at 210mA rather than 240ma.
Hyperbolic A and/or AB would each have crossed far lower than 210ma.
So I submit this mode of operation is similar to HyperA, but 2nd harmonic
of the output currents bent entirely the opposite way.

Though this by far isn't the least efficient JLH design I've seen recently.
I had to lower the load to 32ohms just to exaggerate the graph enough
to take a meaningful picture of what I wanted to show. Not a problem,
just another something to be aware of...

You will need to watch over the thermals, else this design has nothing but
a base current limit to prevent total meltdown... Current gain goes up with
heat, so its uncertain to me if a CCS base current limit alone is sufficient
when you eventually upgrade this to move some real power.
 

Attachments

  • JLH_Ic.gif
    JLH_Ic.gif
    32.9 KB · Views: 533
Last edited:
As I suggested before, turning the lower transistor into a fair current mirror, with thermally coupled pair, would improve things.

Also Keantoken, "dirty" currents mean currents produced by dirt in a crystal, also called as "Leakage" currents.

In this forum, someone stated that all single amplifying elements (tubes, bipolar or FET transistors) should even be considered as differential devices, reacting to a voltage between two inputs. This way to see circtuits greatly simplifies their analysis.

Absolutely right. And NTP or LTP and in such case means one transistor amplifier with an emitter follower in feedback. If to reduce values of feedback resistors due to low output resistance of the output stage we have possibility to eliminate the odd emitter follower in feedback further improving performance.
 
Last edited:
Alright, first of all I was aiming for low parts count, because I didn't want to spend too much time trying to get it together.

As for the output stage, maybe I could do better... And the output offset wanders with supply voltage as well, which can't be good (and yet it still sounds so good). There are ways to filter R2 and R3 so they aren't affected to strongly by supply voltage (might be worth a zener)... They perform an interesting function, which is why I just don't trade them for Jfets any day. I want to explore the sonics of this and see what differences there are.

Another weird thing about this circuit is that net THD is not directly determined by OLG and internal linearity. The NTP pre-buffers add much of their own depending purely on signal voltage. So actually, distortion will be higher for high impedance loads at the same listening levels.

Because of the prebuffer distortion, adding gain reduces distortion dramatically as well. (and then if we used shunt feedback, we could eliminate prebuffer distortion altogether...)

I know there are those here who may disagree with my experiments on increasing distortion to increase listening experience... Well they're just that, experiments. I don't know where I'll end up just yet.

- keantoken
 
Last edited:
It's probably your grounding. If I were you I'd use a couple of 9V batteries in series, with the middle point being 0V, and connect the source ground to this 0V point. Keep all probes off for this test, including DMMs that are not battery operated. If you got rid of the hum thus the next step is obvious.

Can 9V batteries hold up with hundred of mA of current for very long? That might be a while...

- keantoken
 
Actually I had two of those hooked up before I set it up with the batteries.

Problem is, with caps so FRICKEN HUGE, I can't shut it off before it destroys itself if something goes wrong. 😉 (they're working quite well, BTW)

Now that it's on batteries, I hope I can get it working with my soundcard.

- keantoken
 
Good point. Nah, the fuses are too expensive. 😀

I added a inline switch. Also, I don't see how the circuit can blow up by drawing a bunch of current, now that I look at it. If it could, I would have done it by now.

So far nothing has helped the computer noise, oddly enough, except putting a 1uF filter cap in the CCS. That's weird. There's something I don't know. So it looks like it might not be necessary to run off of batteries. The trimmer I'm using has a leg that's not attached to anything, would that cause problems?

- keantoken
 
For me, fixing PSRR took way more than 4 transistors I had originally planned.
Bumping my count to 7. Though two of em probably coulda just been diodes...
I'm not yet sure why yours is so easily worried by power ripple? Nor what is
the least component count to fix it? Overcomplicated power supply defeats
the purpose of a low parts count amp.

Whatever... Compare these currents to the graph of post #46. Not quite as
Hyper as my usual Allison with the Schottkys. Cause I'm fighting against JLH
natural curve wants big end of the hardboiled eggs pointing the other way.

This still does not address the matter of varying emitter currents causing
a non-linearly varying drop in the follower/comparator. I don't think Rush is
the quick fix for that problem? Taylor might be, but implemented differently
(and for different reasons) than I have here.
 

Attachments

Wow, that's a little complex for now...

I'm going to try and reconcile my comments about offset vs. supply voltage. As supply voltage increases, current through R2 and R3 increases, increasing the Vbe of Q7 and Q6, which affects offset compensation. Looking back, I think I overreacted here. But I still think there is room for improvement.

Well, the hum I'm getting through the soundcard is not really hum, it's sharp buzzing. I haven't found any external source of interference, but it appears to be the same frequency as the mains.

What I do know is that there is something producing semi random sharp spikes whenever I turn everything else off and look at the scope screen. It's not the scope. I'm clueless. I'm thinking maybe this interference is what I'm hearing; the cable is a few meters long and an unwitting antenna for powerful bursts of energy. It's there whether or not my computer is on.

- keantoken
 
Wow, that's a little complex for now...

I'm going to try and reconcile my comments about offset vs. supply voltage. As supply voltage increases, current through R2 and R3 increases, increasing the Vbe of Q7 and Q6, which affects offset compensation. Looking back, I think I overreacted here. But I still think there is room for improvement.

Well, the hum I'm getting through the soundcard is not really hum, it's sharp buzzing. I haven't found any external source of interference, but it appears to be the same frequency as the mains.

What I do know is that there is something producing semi random sharp spikes whenever I turn everything else off and look at the scope screen. It's not the scope. I'm clueless. I'm thinking maybe this interference is what I'm hearing; the cable is a few meters long and an unwitting antenna for powerful bursts of energy. It's there whether or not my computer is on.

- keantoken

I'd next run the amp with a simple battery powered cd player, mp3 player, or even :gasp: a walkman 😀 as source. Thus rule out the computer. Is it possible your amp is oscillating?
 
It's possible it's oscillating. The NTP has three oscillation modes. Chaotic pulse, Saw, and Sine. When it fires it fires like an SCR, and with small signal transistors I imagine it can pack some RF. But it only buzzes when I connect the computer, and the input already has a 1nF cap to ground. Removing the cap worsens things. I was listening before on my CD player, no hum.

I just connected the cable to the scope to measure the problem directly.

How about that! a 50mV pk-pk, 100KHz sine wave! Doh! Where did that come from!? (so I imagine perhaps enough RF energy could cause the NTP to go chaotic)

Adjusting my card's ample rate setting causes the signal to change. So this could be normal. But 50mV!? That's not good! (oddly enough, it seems my drivers are only successful at changing the sample rate half the time, because I have to click the option a few times before something happens).

Hm, what to do?

- keantoken
 
Wow, that's a little complex for now...

7 transistors way mo complex than 9 + CCS = 11! I don't think so...

Just ignore my original thought process documented in the drawing, and think
simple JLH. But the CCS is modulated, such that sum of voltage drops across
R1 + R2 is held constant. The two resistors in the middle (Allison) are moved
outboard, but serve the exact same function.

This results in "normal" class A. Neither output current sines are distorted
big endian (JLH) nor little endian (HyperA). Your Allison designs I've seen to
date have all behaved in this, the "normal" Class A way.

Add Schottkys as you've seen me do before, to compute Hyper A curves.


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

JLH without resistors to monitor the output currents? A constant sum base
current yields perfectly complimentary, but somewhat wasteful, big endian
distortions. You have no control what shape the actual output currents
might take (except that they will always compliment), and its guaranteed
to vary with temperature.
 
Last edited:
Status
Not open for further replies.