Help Understanding Circuit (grounded grid I/V)

[dsavitsk]

DIYParadise has published a new grounded grid gain stage for a DAC using a 5842. Here is the schematic



The original page is: http://diyparadise.com/mar06/gg.html

The tube's curves are attached.

There are a couple of things I am having trouble understanding. First, as just a basic of what is going on, am I correct in thinking that since DAC chips are current sources, that by connecting the DAC to the cathode that the current through the tube is changing in the same way that applying a voltage to the grid would change the current?

Second, Yeo notes that the DAC wants to see a low input impedence. Can anyone explain why that would be the case?

Also, the DAC that he uses is a voltage source as it uses a passive I/V. It seems like the DAC should be connected to the grid as a grounded cathode gain stage, or the I/V resistor should be removed to simply inject current into the cathode.

Can anyone make a bit more sense of this to me? I was thinking of doing something similar but using a CCS as the plate load. Would it be better to use a passive I/V and connect to the grid, no I/V and connect to the cathode (that 100uF electrolytic seem slike trouble) or something else?

[EC8010]

Many DACs are made up of an array of constant current sources. The way they're made is that a reference voltage is switched on or off to the base and this sets a voltage across an emitter resistor, thus setting the current. That's all true to a first approximation, but when you look a bit harder, you discover that if you allow VCE to change IC also changes. Bear in mind that the current source that is the MSB has to be set to a precision of the LSB and you soon realise that you can't allow VCE to change. The way to prevent VCE changing is to load it with a very small resistance, ideally zero. Thus, current output DACs normally feed an op-amp I/V convertor.

Feeding the cathode of a grounded-grid stage is quite a low resistance, but it isn't zero. The AC resistance seen looking up into the cathode = (RL+ra)/(u+1) in parallel with Rk. In your example, that's likely to be somewhere around 100R.

[kevinkr]

I built a passive I/V using 10 ohm caddock resistors direct coupled into a common cathode amplifier stage with a gain of about 60, followed by an srpp with a gain of about 14 - this worked pretty well with the BB dacs in my old Assemblage 1 Dac.

It strikes me that this design has a number of problems as EC8010 pointed out, it doesn't look anything like a virtual earth to the dac driving it which is going to cause linearity problems in the dac. I choose 10 ohms in my design because the spec for the particular dac (I forget which one now) indicated its linearity would not be compromised as long as the voltage at the input of the I/V converter did not exceed 22mV.

One of the problems with this design is the input impedance is not constant with frequency, I doubt low frequency performance would be all that good. I also believe there are there are some small dc errors in most dacs that would possibly be magnified by this approach.

Something similar could work dc coupled with a dc servo sensing the cathode and driving the grid negative to force the cathode to ground. A small negative supply is required for both the op-amp and the cathode supply. You could also use a simple current source here to the - supply and put a 10 ohm resistor from cathode to ground to set the input impedance. I actually tried this on the bench and it seemed very promising - I ended up getting a PSA Ultralink II which ended the experiment. Ideally you could use transformer coupling to get the additional gain you need, even a 4:1 step-up (or less) could be employed with the 5842 giving you a source Z of <30K and a gain of about 160. A simple ccs loaded cathode follower could finish it off..

[dsavitsk]

Ok, here is what I am thinking about doing. I have used a 33R I/V resistor with the TDA1541A and it worked pretty well. (I'd like to make this modular so I can pop a board with other DAC's with their I/V resistors in to experiment a bit.) In the past, I just DC coupled this to the + input on an opamp with a gain of 20 or so with no problems.

In the design below, would I need to add a small coupling cap between the I/V resistor and the input resistor, or can the dac just go right into the grid of the tube? If so, could I avoid the coupling cap with a transformer I/V?

The tube seems pretty well suited for this -- high enough gain, low enough plate resistance (my preamp has a 50K input, so the 1.7K output should not cause trouble here.)

[pengboon]

Am I understanding correctly by saying the DAC prefers a low input impedance preamp?

So, if this DAC was hooked up to a CC tube preamp, say a 12B4, what would be the effect? I think some may be using this in their system...

thanks...

[kevinkr]

Yes, that looks like a pretty good implementation, you don't need a coupling capacitor between the dac output and the grid of the 5842.
Depending on the dac you may want a smaller I/V conversion resistor than the 33 ohm you are currently employing.
Some data sheets will publish the maximum voltage that can appear at any frequency on the I output of the dac before errors start to affect the monotonicity of the dac output. In the case of the BB it was just a little over 22mV..

[G]

I am thinking about trying this also as I have a sleeve of 5842 in the handy dandy tube caddy collecting dust. Have you had any luck with this project dsavitsk?

[dsavitsk]

Originally posted by G
I am thinking about trying this also as I have a sleeve of 5842 in the handy dandy tube caddy collecting dust. Have you had any luck with this project dsavitsk?

Lots of luck. I didn't end up using a 5842, but I have tried both a 6n6p and a 5687 (with both a tda1541a and a ad1865 based dac both with passive i/v) and am happy with both. Even the opa637 based opamp gai stage sounds pretty good. Schematics and some details (though not up to date) at http://www.ecp.cc/modular_dac.html

[G]

Originally posted by kevinkr
I built a passive I/V using 10 ohm caddock resistors direct coupled into a common cathode amplifier stage with a gain of about 60, followed by an srpp with a gain of about 14 - this worked pretty well with the BB dacs in my old Assemblage 1 Dac.

It strikes me that this design has a number of problems as EC8010 pointed out, it doesn't look anything like a virtual earth to the dac driving it which is going to cause linearity problems in the dac. I choose 10 ohms in my design because the spec for the particular dac (I forget which one now) indicated its linearity would not be compromised as long as the voltage at the input of the I/V converter did not exceed 22mV.

One of the problems with this design is the input impedance is not constant with frequency, I doubt low frequency performance would be all that good. I also believe there are there are some small dc errors in most dacs that would possibly be magnified by this approach.

Something similar could work dc coupled with a dc servo sensing the cathode and driving the grid negative to force the cathode to ground. A small negative supply is required for both the op-amp and the cathode supply. You could also use a simple current source here to the - supply and put a 10 ohm resistor from cathode to ground to set the input impedance. I actually tried this on the bench and it seemed very promising - I ended up getting a PSA Ultralink II which ended the experiment. Ideally you could use transformer coupling to get the additional gain you need, even a 4:1 step-up (or less) could be employed with the 5842 giving you a source Z of <30K and a gain of about 160. A simple ccs loaded cathode follower could finish it off..

Sorry for the ignorant question but since humans don't hear linearly does it matter if the load causes linearity problems?