• 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.

Differential Amp CCS Tail SVCS Plate - Noise Issues

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
A closer look at the circuit reveals a definite problem - and it may well explain what Ian's seeing.

The 8233 data shows triode-mode characteristic at the nominal 125V on the anode. At this voltage, Ian's desired 30mA current is achieved with a bias of about -3,5V. At 110V, it will be -3V, or even closer to ground.

This bias voltage has to establish the working voltage of the LM317. In current-source connexion, the LM317 requires 1,2V to satisfy the reference, plus the dropout voltage of the regulator parts. the 1978 National data shows a typical value of 1,5V at 20mA (25 deg C). So the minimum operating voltage for the CCS will be about 2,7V - with a strong temperature dependence (needs more volts when cold).

As you increase the 8233 anode voltage from say, 80V, you begin by offering the LM317 only about 2V to work from. In this condition, it will either draw no current, or turn ON hard. It will certainly not be regulating the current! At a certain anode voltage, the bias required by the 8233 will meet the minimum required voltage for CCS operation, and it will start up. I think this is what you are seeing as you turn up the voltage. It may just be that the LM317 is unstable in the region of not-quite-enough operating voltage.
 
Last edited:
I am still not convinced by the idea of having a CCS in the cathode circuit, when one has a (piecewise, at least) CCS in the anode.

The easiest way to confirm or deny my claim: slide a 100R 0,5W in place of each LM317.

If you are committed to the CCS in the cathode though, you'll need one with the compliance extending all the way to within 1,5V of ground.

It would also be an advantage to be able to trim the working bandwidth of the cathode circuit, so that it does not provoke instability in your (very) high gm pentode.

Here's my take. The circuit attached gives a reasonably balanced and tracking pair of currents, especially if you epoxy the transistors together.

C2 and C3 provide poles to adjust, so that you can damp down oscillations provoked elsewhere. These can be raised to 220n if desired, MKP stacked film parts, please!

Best of all, the compliance extends down to +1,4V; so you can easily bias the 8233 within the optimal operating range of this CCS.
 

Attachments

  • LVCCS2PP.jpeg
    LVCCS2PP.jpeg
    51.6 KB · Views: 268
I am still not convinced by the idea of having a CCS in the cathode circuit, when one has a (piecewise, at least) CCS in the anode.

The easiest way to confirm or deny my claim: slide a 100R 0,5W in place of each LM317.

If you are committed to the CCS in the cathode though, you'll need one with the compliance extending all the way to within 1,5V of ground.

It would also be an advantage to be able to trim the working bandwidth of the cathode circuit, so that it does not provoke instability in your (very) high gm pentode.

Here's my take. The circuit attached gives a reasonably balanced and tracking pair of currents, especially if you epoxy the transistors together.

C2 and C3 provide poles to adjust, so that you can damp down oscillations provoked elsewhere. These can be raised to 220n if desired, MKP stacked film parts, please!

Best of all, the compliance extends down to +1,4V; so you can easily bias the 8233 within the optimal operating range of this CCS.

You could of course generate a small negative rail instead. If you tried this then something like +6v referenced to the LL1544 would be fine, simply rectify a spare heater tap and regulate it. This should be very easy to test with a little 9V battery between the CT and ground.

It still doesn't cure the problem of the LM317 been a poor CCS in this position.

Shoog
 
Last edited:
I've been playing with this circuit now for a few days. It can sound very good but there is a noise issue I'm not sure how to approach. When the "right side" SVCS board is set to anything above 108 VDC there is a white noise (pfffffff) audible on the speaker. (The noise off-on knee is sharp, over something like a volt or two) If the other SVCS is taken up above about 120 that noise increases by something like a third. If I take the right hand plate on its own up to about 180VDC the noise then stops.

Thinking there might be some kind of current source interaction between the SVCS and the LM317 I have tried bypassing the SVCS loads with 180K and also 100K resistors. I've also tried bypassing the driver tubes with a 180K resistor from plate to ground. I think the noise is lowered a little in each case but by hardly enough to justify leaving the resistors in.

The amp sounds great with the driver plates below 108V but the operating point can hardly be dictated by the limits of a problem. I believe something is oscillating but where and why I don't know.

I also tried resistors for plate loads. No hiss but the sound isn't nearly as good as with the SVCS. Can anyone offer suggestions?

The SVCS board circuits are as in the Schematic posted here but with Q1 = ST P3NK80Z , Q2 = FQU11P06TU , Q3 = MJE5731A

Also not shown on my drawing both 8233's have 100R stoppers on G1

I don't mean to be harsh, but this a bit of a mess...

Please, only ONE CCS in the LTP. You have way too many parts in there. It doesn't appear to be a true LTP anyway, since the cathodes are not DC connected. Replace the anode loads on the input pair with just two resistors. The output tubes can have one CCS with the 2 degeneration resistors, no problem, but two CCS's will cause problems.

Another problem is that the CCS in the LTP do not have enough of a voltage drop across them to function, since you've connected them to ground and the grids are grounded as well, unless you have a huge bias on those input tubes.
 
Last edited:
I believe you are trying to do something like this. I don't have spice models for your particular tubes, so this is just an example of possible schematic. If you use CCS's, you generally need a B- supply, depending on what kind of CCS you are using. Hope this helps. The only way to control the gain in this kind of circuit that I know of is to change the degeneration resistors in the first LTP. They have to be matched, of course. You may have to do things like capacitor couple the input stage to the output stage, or other things that I have not shown. Other details that I have not shown are things like how to deal with DHT's and CCS's as well as degeneration resistors. I don't have a clue about that. That's for someone else to work out. I just wanted to show the basic concept here. This is not a working schematic, in other words.
 

Attachments

  • Tube push pull power.JPG
    Tube push pull power.JPG
    58.9 KB · Views: 204
Last edited:
I believe you are trying to do something like this. I don't have spice models for your particular tubes, so this is just an example of possible schematic. If you use CCS's, you generally need a B- supply, depending on what kind of CCS you are using. Hope this helps. The only way to control the gain in this kind of circuit that I know of is to change the degeneration resistors in the first LTP. They have to be matched, of course. You may have to do things like capacitor couple the input stage to the output stage, or other things that I have not shown.

There are advantages to using the bias arrangement on the input LTP - but much more so on the output stage. It creates zero DC in the output transformer and allows for self adjustment to compensate for aging of the driver tubes and voltage drift on the plates. Having the cap coupled differential pair on the output also allows for a single CCS on the input as the output will automatically self adjust for differences in the driver tubes plate voltages.
A DC coupled amp with no DC in the output transformer and no adjustment over the lifetime of the amp is worth something in my book. The cap coupled differential stage should work just about as well as a single CCS version - if the CCS is stiff enough and the caps are adequately bypassed.

You may choose other compromises - but the advantages to this arrangement are many and varied and not all obvious at first glance. However the implementation attempted here misses most of the advantages possible. The arrangement you suggested is inherently unstable over time which doesn't seem a very good solution at all.

Shoog
 
Last edited:
There are advantages to using the bias arrangement on the input LTP - but much more so on the output stage. It creates zero DC in the output transformer and allows for self adjustment to compensate for aging of the driver tubes and voltage drift on the plates. Having the cap coupled differential pair on the output also allows for a single CCS on the input as the output will automatically self adjust for differences in the driver tubes plate voltages.
A DC coupled amp with no DC in the output transformer and no adjustment over the lifetime of the amp is worth something in my book. The cap coupled differential stage should work just about as well as a single CCS version - if the CCS is stiff enough and the caps are adequately bypassed.

You may choose other compromises - but the advantages to this arrangement are many and varied and not all obvious at first glance. However the implementation attempted here misses most of the advantages possible.

Shoog

I don't have a clue what you are talking about without a schematic. I tried to help the OP. Obviously, you can choose to create something different. Please go ahead and draw something up and post it like I did. I did my part and now I'm done.
 
I build input phase splitters all the time with a CCS in the tail and gyrators on the plates and they don't oscillate. That by itself is not the problem. If you have a standard OPT it can handle a little DC and your driver voltage adjustment using the gyrators will do the job.

You do need to bias the tail CCS up higher, just put a few volts +DC on the center tap of your input transformer instead of grounding it. In fact, this is a good solution and keeps cathode current out of the bias supply.

I also must throw my hat in and say there must be a better way to do the tail CCS than LM317. Also I don't think the DC balance on the input tubes is needed at all. The 2 input tubes can run at somewhat different DC currents no problem. I'd just put in a DN2540 or 10M45 with about 15V or more across it and be done.

I spoke with this guy named John at some audio show who builds 300B parafeed amps and he said the same thing about needing the 900V part in the output tail...

PS another thought. The input transformer is already a great phase splitter. You could also separate the cathode circuits of the 2 input tubes. There may be a sonic benefit in getting rid of the cathode junk altogether under those high slope tubes and fixed biasing them through the transformer. just saying since we're opening up the box a little...
 
Last edited:
Member
Joined 2008
Paid Member
A closer look at the circuit reveals a definite problem - and it may well explain what Ian's seeing.

The 8233 data shows triode-mode characteristic at the nominal 125V on the anode. At this voltage, Ian's desired 30mA current is achieved with a bias of about -3,5V. At 110V, it will be -3V, or even closer to ground.

This bias voltage has to establish the working voltage of the LM317. In current-source connexion, the LM317 requires 1,2V to satisfy the reference, plus the dropout voltage of the regulator parts. the 1978 National data shows a typical value of 1,5V at 20mA (25 deg C). So the minimum operating voltage for the CCS will be about 2,7V - with a strong temperature dependence (needs more volts when cold).

As you increase the 8233 anode voltage from say, 80V, you begin by offering the LM317 only about 2V to work from. In this condition, it will either draw no current, or turn ON hard. It will certainly not be regulating the current! At a certain anode voltage, the bias required by the 8233 will meet the minimum required voltage for CCS operation, and it will start up. I think this is what you are seeing as you turn up the voltage. It may just be that the LM317 is unstable in the region of not-quite-enough operating voltage.


Rod,
I have separate power for the filaments and heaters so startup is governed by the RC circuit in the SVCS.
When I first powered this up I started the driver tubes off with Vp ~150 Ip 30mA, the idea being to get compliance on the 317. Just barely enough I know but enough I think. In your comment above you say that above Vp80VDC on the 8233 the 317 wil only have 2V to work from but it's actually not the case. The voltage across the 317 rises as VP does and at Vp150V measures 4.7 on the breadboard.

Also, I tried trimming grid leads and adding more R to the 8233 grid stoppers but they had no effect.
Thanks for advice on proximity. I didn't know that and will watch for it in future.
I'll also try your other suggestions as I can but I'm getting overwhelmed by everything here and it will take a while to try everybody's suggestions.
 
Member
Joined 2008
Paid Member
I build input phase splitters all the time with a CCS in the tail and gyrators on the plates and they don't oscillate. That by itself is not the problem. If you have a standard OPT it can handle a little DC and your driver voltage adjustment using the gyrators will do the job.

Good to hear. Thanks for the encouragement. The experiment really is a lot of fun . . . .


I also must throw my hat in and say there must be a better way to do the tail CCS than LM317. Also I don't think the DC balance on the input tubes is needed at all. The 2 input tubes can run at somewhat different DC currents no problem. I'd just put in a DN2540 or 10M45 with about 15V or more across it and be done.

I've had the same thought many times. But this started by seeing Shoog's post and thinking the idea was good to try. I'll push it a little farther before giving up. The circuit as is sounds very good and makes everybody who walks in while it's on start tapping and moving.


I spoke with this guy named John at some audio show who builds 300B parafeed amps and he said the same thing about needing the 900V part in the output tail...

Some guy named John? Hmmm, not sure if you're bringing in backup or throwing me in with the audiophile folklorist society. :)


PS another thought. The input transformer is already a great phase splitter. You could also separate the cathode circuits of the 2 input tubes. There may be a sonic benefit in getting rid of the cathode junk altogether under those high slope tubes and fixed biasing them through the transformer. just saying since we're opening up the box a little...

I've been thinking this from the start.

Also, I did try putting a scope on it for a bit last night but time was limited. I don't have a working differential probe nor 100X regulars so put one probe on each 8233 cathode, inverted channel 2 and added. I got a 100MHz. signal on the scope and was excited until I turned the amp off and found the signal still there. I'll try more tonight.
 
Member
Joined 2008
Paid Member
I worry a little that while I'm trying all these different things my lack of knowledge has me getting results unrelated to my experiment because of mistakes made in their implementation or layout.

Anyway, not being confident in my scoping skills, last night I decided first to put 150R//680uF under the output tail CCS and take the input trans CT to the tap, putting about 20VDC across the LM317. This got a little more white noise and now with a buzz component.

Then I took out the capacitors from the driver cathode circuit. Lots more dirty white and no music material getting through at all.

Then ran the input trans CT back to ground and replaced the 317s with 150R in each cathode, first un-bypassed and then with 680uF across each. After warming up for about an hour I muted the computer feeding it and the speaker was silent.
Almost always as I get close to the voltage of one while trimming the other but not limited to then.

The other question in my mind about the original noise level was that at one point I turned down the variac feeding the power transformer until the noise stopped, waited a few seconds and then turned it back up. It took somewhere between 5 and 10 seconds for the noise to come back on - ramping up over about 3 seconds. It made me wonder if there is a thermal issue somewhere I can't see. . . . . . and so on. (eg. The centre transistor on the SVCS (FQU11P06) doesn't have a heatsink as it sees less than 10V at 30Ma and measures 39 degrees C - [and to boot it's an I2Pak and I couldn't seem to find a sink for it anywhere])

All these results are very interesting as they seem to substantiate what some have been saying , yet I'm convinced there are ways to make these things work and want to persue them.

Shoog's statement that the capped cathode arrangement would be better for the output stage and would guarantee balance is of interest because I see the balanced floating around though moving in a small range between 1 and 5 milliamps.

I also question the present output tail CCS stability and wonder if there are any suggestions for a stable circuit that would take 200+VDC at 150mA.

Also, now that the breadboard is already there, I'd like to try putting an anti-triode on on one end of the primary at some point.

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