• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

DC coupled Sand 'n Glass with Current Feedback

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.

Attachments

  • Photo on 12-03-12 at 6.17 PM.jpg
    Photo on 12-03-12 at 6.17 PM.jpg
    28.2 KB · Views: 197
Going by the circuit in the first post, I'm a bit confused as to how your feedback scheme works. With the unbypassed ccs in the cathodes there is basically no ac path to ground making the gain set resistors in each cathode moot, making the feedback signal work against the tubes internal resistance as the only path to ac ground. Also if you were to feed each grid from a balanced signal, +g1 -g2 the cathodes would fight each other and do some other unknown things with the feedback from the outputs also being out of phase.

Perhaps something like in the attached would be more "correct". S1 changes between balanced and single ended input.
 

Attachments

  • sand&glassfft.jpg
    sand&glassfft.jpg
    20.6 KB · Views: 177
  • sand&glass.jpg
    sand&glass.jpg
    71.3 KB · Views: 192
Last edited:
Hi Bigun,
yup, that's what I understood.
As long as I keep it with single stage feedback only, then one diamond buffer would be sufficient.
If I intend to do a fully symmetric feeback across input stage and VAS then I would need to bootstrap also the second VAS.
I like the idea of bootstrapped VAS.
Regarding the input, I have concerns regarding common mode rejection.


Hi Jerlowu,
the difference between your feedback and mine is that you are doing an unbalanced voltage feedback, while I did a balanced current feedback.
Basically both should work.
Spectrum of the distortion does not look promising in your sim.
Is reality better or worse?

Current feedback:
In case of having cathode resistors much smaller than 1/g of the tube we could simplify gain to 1+Rfb/Rcath. In fact we do not have tubes with such high transconductance and consequently gain is smaller.
The advantage of the symmetric current feedback is large bandwidth and low noise.
Equivalent input noise is dominated by the tube vs. domination of 100k in your case.

I have to check on your comment regarding fighting feedback tails.
So far I do not expect this from theory - still not checked on the proto.

...but first I have to repair the car... :mad: :mad:
 
Triodes

Over the last week my triodes arrived.

1. 6DJ8
Ordered at one of the largest DIY-stores in Germany.
The tubes were just some unspecified tubes.
Not even a copy of a 6DJ8. Just junk with completely wrong parameters.

2. E88CC (JJ)
Ordered at Tube-town.
The tubes perform more or less like they should.
Mechanically they tend to vibrate, when you knock at them.

3. 6922 (Sylvania)
Ordered at Tube-town.
Also these tubes perform more or less like they should.
No tendency to vibrate. Slightly higher transconductance than the JJs.
Slightly preferable distortion in the circuit of posting #1 vs the JJs ( No listening test so far.).
 

Attachments

  • Triodes.JPG
    Triodes.JPG
    79.5 KB · Views: 128
Voltag Feedback

Besides ongoing work on the current feedback circuit, I had a look to an arrangement with voltage feedback. Similar as proposed by jerluwo.

Slightly different from jerluwo I have focussed on the non inverting input arrangement, but settled symmetric outputs.
Nevertheless I have to thank jerlowu! Without his proposal I might not have had a look to any arrangement with voltage feedback, while it is really worth to be considered!

Step response is fine.
Distortions also are looking OK. Nice thing is that the increase of distortion at high output levels is moderate. Spectrum is more pleasant than in the simulation of jerluwo.

First the schematic, then the distortion measurements at 5Vp, 10Vp, 20Vp and 40Vp. The dual triode was the 6922 from Sylvania.
 

Attachments

  • 40Vp_60V_Vfeedb_6922.JPG
    40Vp_60V_Vfeedb_6922.JPG
    92.1 KB · Views: 31
  • 20Vp_60V_Vfeedb_6922.JPG
    20Vp_60V_Vfeedb_6922.JPG
    92.5 KB · Views: 32
  • 10Vp_60V_Vfeedb_6922.JPG
    10Vp_60V_Vfeedb_6922.JPG
    91.3 KB · Views: 31
  • 5Vp_60V_Vfeedb_6922.JPG
    5Vp_60V_Vfeedb_6922.JPG
    93.1 KB · Views: 73
  • HybVfdbSchem.JPG
    HybVfdbSchem.JPG
    105.5 KB · Views: 133
Triodes are only linear with a high impedance load.

Combined with the bootstrapping proposal of bigun there is a promising arrangement.
Well, I am not going to bootstrap the VAS output, because this would cause my VAS bias to become a function of the supply voltage.
But it should work fine to bootstrap both anode tails and provide by this a high impedance load for the triodes.
 

Attachments

  • HybBtstrpSchem.JPG
    HybBtstrpSchem.JPG
    113.8 KB · Views: 77
In any case the distortion figures are not bad at all, so I started to dig into other properties and easily found that the IMD were higher than desired.
The reason for the high IMD was the bipolar VAS.
Straight cascoding did not help.
Changing to MosFets did help a lot.
Changing to MosFets and cascoding with BJTs improved things by 30db.
Furtheron this configuration is pretty fast. The proto is showing a slew rate of 170V/us and a blameless step response without overshoot.

Attached the new schematic.
2nd picture is the step response, probes were again 10X.
3rd picture is the overview of THD and IMD at different levels.
4th picture is the spectrum of the THD at 20Vp.
 

Attachments

  • HybCascSchem.JPG
    HybCascSchem.JPG
    112.1 KB · Views: 206
  • Step.JPG
    Step.JPG
    107.3 KB · Views: 207
  • Distortion_overview.JPG
    Distortion_overview.JPG
    36.4 KB · Views: 178
  • 20Vp_THD_casc.JPG
    20Vp_THD_casc.JPG
    82.2 KB · Views: 188
After going back and looking at the cathode feedback your using I had missed at first that the ac feedback currents were going from vas to vas. Now the issue that concerns me with this is the dc feedback. Having one vas shifting its offset will have an effect on both cathodes changing the offset of both vas outputs. It may not really be a big issue in the real world circuit, but, if your going to use a dc coupled output to the speaker, this would be something to watch.
I would agree that the distortion is mainly the vas part as you found. The tubes are only swing a very small amount of ac voltage and current and contribute only a fraction of the total distortion. Their limited gain compared to a bjt diff. pair limits the amount of correction they can enforce on the vas, making it necessary to make the vas as linear as possible before feedback to get distortion down. This low gain/feedback combination probably plays a big part in the nice step response as well.
I do hope you try a bridge output on this circuit. That second vas doing nothing but feedback gives me the willys. Just a personal thing I guess.
 
...keeping one tail of the VAS unused except for feedback - yes that's sort of ignorance.. :eek:
Feeling the same.

Regarding DC. This circuit is definitely more demanding regarding adjustment than most standard solutions. I did not put the adjustment details into the schematic.
You have to adjust two parameters.
a) Bringing both outputs to the same value.
b) Bringing them to zero vs GND.

In fact I planned to use this circuit to drive a high level input of a class D amp.
But since some days I am thinking about keeping the class D system in pure sand and enhancing this triode circuit to a full power amp.
It should be possible to run the VAS at higher currents and put a heavy class A diamond buffer or a class A MosFet output stage....
...could be even bridged.

Regarding a bridged version I would expect that K2 will be mostly canceled, but not K3.
Do you guess the same?
 
For curiosity I simply build a differential measurement amp with an OP amp.
It seems like the limits of this measurement amp are well below the needs.
At least the measurement results were stable no matter if I used a 1/2 of an OPA2134 or TL072 or NE5534.

Let's have a look to the measurement results.
The measurement points were the VAS tails.
Measurements of single tails was done by using the non inverting input of the measurement amp.
Tail 1 is the tail that feeds the diamond buffer.
K2 on both tails are not fully identical.
Consequently in differential measurement they do not fully cancel, but go down by 15db (vs. Tail1).
That's fitting to what one would expect.

K3 do not reduce each other, but seem to increase.
I would have expected a similar levels like the separate tails.
:scratch:
 

Attachments

  • Comparison.JPG
    Comparison.JPG
    74 KB · Views: 141
Sorry, I do not critize (do you mean criticize ?). I was searching for something else.

Search function leads you to old threads most of the time and it is useful (the search
function will NOT show you future results - not sure why). This forum is a (short)
history machine.

What do you want to tell me? It is always useful for beginners if things are made up
clearly. The forum, as I understand it, can be a learning machine. Some (few) people
want to learn even six years later. Others may be new to this thread.

If you think it is essential that I stay out, please let me know. Thanks.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.