Bias Blues

[Ten Over]

I agree that 220K is way to big. Just because I can, I mocked it up in LTspice and found that he should still be getting between -3V and -5V bias, though it takes ~30 seconds to get there. If sepulchre is really getting ZERO volts then there must be a wiring error.

That's amazing because I reckon'd 6V at the head of a 15K/25K/47K voltage divider which would give 5V at one end of the 25K and 3.2V at the other. How'd I do that?

That zero volt thing crossed my mind, also.

[Tony Bones]

I agree that 220K is way to big. Just because I can, I mocked it up in LTspice and found that he should still be getting between -3V and -5V bias, though it takes ~30 seconds to get there. If sepulchre is really getting ZERO volts then there must be a wiring error.

That's amazing because I reckon'd 6V at the head of a 15K/25K/47K voltage divider which would give 5V at one end of the 25K and 3.2V at the other. How'd I do that?

That zero volt thing crossed my mind, also.

My god! That's millions of calculations in your head!

If I wasn't sitting in front of a computer I suppose I could have assumed the first filter cap charged to a full 55V, then there's just a DC voltage divider... But I like looking at the pretty graphs on the computer screen.

[Ten Over]

My god! That's millions of calculations in your head!

Naw. Only one table, an assumption, and three calculations.

[Ten Over]

...I think the difference between the two is probably that in the original setup, the cap is being charged by the voltage divider of 87K/307K, but also through the Thevenin equivalent series resistance of 87K || 220K = 62K, so the charging pulse is spread out too wide and the average voltage goes down, as well as the charging voltage being divided down.

I'm not fully following you on that, but if you want to think in terms of voltage dividers, you could model the capacitor as a resistor during the charge portion. I mean the capacitor draws current during charging and that has the same effect on the voltage drop across the 220K resistor as a resistance in parallel with the 87K to ground.

[R.G.]

Let me try to explain better.

We can model the secondary of the transformer as an ideal voltage source with wiring resistance between it and the rectifier/cap. The diode has some resistance and an offset voltage of about 0.5V (for a simple model) and the wires have some resistance. The cap is loaded by the conglomerate 87K.

If the cap starts at some voltage across it, it runs down by the 87K/Cap time constant. It runs down until the incoming AC voltage exceeds the voltage on the cap and the diode threshold. Then the cap charges at a rate determined by the voltage difference between the incoming voltage and the cap voltage divided by the sum of the transformer wire resistance, the diode resistance, and the wiring resistance.

When the incoming AC voltage peaks and starts down, it rapidly decreases until the diode is no longer conducting, and then the cap starts its exponential decay with the 87K load. So all the charging happens on the slope of the incoming AC that forward biases the diode, then quits until the voltage again forward biases the diode, one AC cycle later. The cap charges in a blip of current on the front slope of the AC, and stops near the peak.

At least that's what happens with the sum of transformer wire resistance, diode resistance, and wiring resistance being "low". And this would be the case if the cap were connected right after the diode. If the cap is after that 220K, the "wiring resistance" becomes relatively huge. That changes things.

First, since the transformer+diode+wiring is no longer trivially small compared to the 87K load, you can't ignore the 87K load as a voltage divider while charging. With the 220k in series with the cap, it charges from a Thevenin equivalent voltage of the 87k/(220k+87k) divider times the incoming AC voltage, and it charges through the parallel equivalent of the 87K and 220K, about 68K. So the cap charges much more slowly when the voltages let it charge. In fact, it charges at a time constant of 47uF and 68K, or 3+ seconds. This means it will never fully charge - in fact, it will be a continuous dribble over most of the half-cycle the diode lets through.

Moving the cap to right after the diode lets the cap charge to the peak incoming voltage minus a diode drop. Then it runs down by the 220K+87K resistance, with a time constant of 14 seconds. So the cap stays up near the peak voltage (minus a little ripple) all the time.

The position of the cap before or after the 220K has a huge effect on the capacitor's DC voltage, and on the amount of filtering it can do. Note that putting the 220K before the diode is the same as putting it after the diode.

Edit: I wanted to be sure I wasn't making stuff up in my head, so I dumped this into the simulator. With the 220K ahead of the cap, the voltage at the top of the 25K pot was about -14V and 50mV of ripple; with the 220K after the cap, it was -38V and 10mV of ripple.

[Ten Over]

If the cap is after that 220K, the "wiring resistance" becomes relatively huge. That changes things.

First, since the transformer+diode+wiring is no longer trivially small compared to the 87K load, you can't ignore the 87K load as a voltage divider while charging. With the 220k in series with the cap, it charges from a Thevenin equivalent voltage of the 87k/(220k+87k) divider times the incoming AC voltage, and it charges through the parallel equivalent of the 87K and 220K, about 68K. So the cap charges much more slowly when the voltages let it charge. In fact, it charges at a time constant of 47uF and 68K, or 3+ seconds. This means it will never fully charge - in fact, it will be a continuous dribble over most of the half-cycle the diode lets through.

Moving the cap to right after the diode lets the cap charge to the peak incoming voltage minus a diode drop. Then it runs down by the 220K+87K resistance, with a time constant of 14 seconds. So the cap stays up near the peak voltage (minus a little ripple) all the time.

The position of the cap before or after the 220K has a huge effect on the capacitor's DC voltage, and on the amount of filtering it can do. Note that putting the 220K before the diode is the same as putting it after the diode.

Edit: I wanted to be sure I wasn't making stuff up in my head, so I dumped this into the simulator. With the 220K ahead of the cap, the voltage at the top of the 25K pot was about -14V and 50mV of ripple; with the 220K after the cap, it was -38V and 10mV of ripple.

Let's run with the Thevenin equivalent for a minute. We have a 55Vpeak voltage source with a 220k resistor in series which connects at the distal end to an 87K resistor in parallel with a 47uF capacitor. The capacitor is the load, so we remove it to get the Thevenin voltage of 15.6V. Now we remove the voltage source to get the Thevenin resistance of 62.3K. So we have a 15.6V voltage source charging a 47uF capacitor through a 62.3K resistor. From this I know that the capacitor will charge from 0V to 9.9V in 2.9 seconds when the voltage source is pure DC. This is right in there at useless to me. I need to know the voltage when the capacitor starts to charge and the voltage when the capacitor starts to discharge. From this I can get the average DC voltage, the ripple voltage, and the time during which the current spike exists.

Tony Bones got -5V at the top of the 25K pot with his simulator, but you got -14V. -14V isn't very likely because you would need 16.9V at the top of 15K and we're only feeding it 15.6V according to the Thevenin equivalent circuit. -38V isn't very likely with the capacitor connected to the diode, either. 39Vrms x 1.35 is 52.7V and that is what I would expect.

[sepulchre]

Okay I FINALLY got logged in. The system rejected me since my first post. That's why I haven't replied.

Thank all you guys for your answers. First, I replaced the mistaken 220K resistor with a 1.5K. MUCH better! So yes to 9000 +ones.

However, this amp has some other issues that I am working out now that I won't bring up. Shouldn't be to difficult to find.

But, the rectifier produces around 470 odd volts - too much for the 5881s I'm using. Can I put a Zener between the negative end of the bridge and ground to reduce that somewhat. The tubes can handle up to 400v so is that not too far to go?

Thanks again for all your help!

Ken Graves
Activax Amplification

[R.G.]

Tony Bones got -5V at the top of the 25K pot with his simulator, but you got -14V. -14V isn't very likely because you would need 16.9V at the top of 15K and we're only feeding it 15.6V according to the Thevenin equivalent circuit. -38V isn't very likely with the capacitor connected to the diode, either. 39Vrms x 1.35 is 52.7V and that is what I would expect.

I'll redo the sim and post some of the results. If I remember correctly, I wasn't using 39vrms for the ac, as I was after demonstrating the principle. I was away from computers the last half of the day, so it'll have to wait til tomorrow. I'll post some pics tomorrow.

We can't have two simulators with different answers, can we? That would be horrible!

[Ten Over]

But, the rectifier produces around 470 odd volts - too much for the 5881s I'm using.

Is there any load (e.g. tubes) when you measure that 470V? What is your wall voltage? What kind of 5881's do you have?

[pompeiisneaks]

Okay I FINALLY got logged in. The system rejected me since my first post. That's why I haven't replied.

Thank all you guys for your answers. First, I replaced the mistaken 220K resistor with a 1.5K. MUCH better! So yes to 9000 +ones.

However, this amp has some other issues that I am working out now that I won't bring up. Shouldn't be to difficult to find.

But, the rectifier produces around 470 odd volts - too much for the 5881s I'm using. Can I put a Zener between the negative end of the bridge and ground to reduce that somewhat. The tubes can handle up to 400v so is that not too far to go?

Thanks again for all your help!

Ken Graves
Activax Amplification

Here's a great thread on zener voltage dropping, yes you can.

https://ampgarage.com/forum/viewtopic.php?t=3782

~Phil

[R.G.]

I did some sim studies, this time using 39Vrms for a source.
I just did two versions of the circuit, and scoped them both in parallel, as below.

Bias Study 1.JPG

First one shows the difference in capacitor attachment, whether before or after a 220K dropping resistor. Top trace (c in the voltage list) is attachment before the 220K resistor, second one down is d, the voltage at the 25K; 52.7V, and 12.358 out at the 14 second mark. The third trace down is the voltage at the first capacitor when connected after the dropping resistor, and fourth is the voltage at the top of the 25K, 5.2V and 4.255V at 14 seconds. What I took from this is that the 220K severely limits the charging on the first filter cap.
The "instant" value of voltage on the fast charge looked funny, so I changed the time scale, as below. Indeed, it does do the rachet-up thing on the first few cycles after turn-on.

Bias Study 2.JPG

The suspicion on the thread was that the 220K should be vastly smaller. I changed that to 220 ohms, and got the pic below.

Bias Study 3.JPG

Sure enough, now the raw bias voltage on the first cap and the voltage at the top of the 25K now get much larger than when choked off by the 220K. But it's still not as big as the cap-before-resistor case. My theory was that the resistor was choking off charging pulses, stretching them out and not letting the cap charge as quickly and hence not to as near the peak of the incoming AC. So I put current probes on the two diodes; sure enough, the probe on the resistor-before-cap line was much wider, indicating a longer charging time.

Bias Study 4.JPG

I increased the series resistor from 220 to 4.7K to see the difference that makes.

Bias Study 5.JPG

Sure enough, the cap-after-resistor case took even longer to charge, but there was no effect on the charging pips for cap-before-resistor.

So I think the sim shows what the theory predicts: diode-resistor-cap setups charge faster when the resistor is low. In this case, putting the cap right after the diode charges the cap to a usable bias voltage fastest. This lets one get bias on the tubes FAST, and tinker it with resistor dividers to the necessary level.

[Ten Over]

Can you align the time to a peak 30 cycles or so down the line for the 4.7K sim?

[R.G.]

Maybe. I'll go try to make that work.

[sepulchre]

But, the rectifier produces around 470 odd volts - too much for the 5881s I'm using.

Is there any load (e.g. tubes) when you measure that 470V? What is your wall voltage? What kind of 5881's do you have?

[sepulchre]

Sorry, didn't get to finish that last post.

I get a fairly steady 120v from the wall sockets. The 470v main I quoted was with so EL34s in it because before I fixed the Bias resistor I saw the 5881s (Tung Sol, incidentally) get a Tiny bit of red plating going on and turned it off IMMEDIATELY, of course. Now with 200R in the bias spot I'm getting around 430v with the 5881s, still too much. Especially since the 400v max is for a triode configuration.

I've been building smaller amps with a pair of EL84s self biased for the last couple of years which have been selling pretty well. So I've kind of gotten away from some of the theory I need here. So I need to read up again on differences between class AB1 and class AB2. In either case the data sheet says the max plate v is 360 so I'm going to give that thread a good going over. Thanks for that pompeiisneaks.

The rest of the amp has other issues as well. I built it while on pain killers after neck fusion (duh), but I think it'll be a good one when I'm done.

Thanks again everyone. Btw, the bias analysis is super!