Don't Regulate Bias Voltage!

[David McGown]

I know this is understood by the experienced DIY'ers, but was trying to explain to a relative NOOB why using a LM417HV in a bias circuit was a bad idea. The attached figure I drew up illustrates the problem. Please ignore the bad operating point I chose (above the plate dissipation limit), it was purely for purposes of discussion.

In this particular example, an increase in line voltage from 117V to 125V (a 7% change), produces a 48% increase in plate current with a regulated fixed bias, whereas with a filtered bias supply (floating with voltage changes), there is an 18% increase in plate current. Obviously, you would not expect to see a 7% variance in voltage, but is not unusual to see fluctuations of 2 or 3 volts depending on the time of day and load. Regulated bias will result is a pretty significant change in operating characteristics.

So don't regulate bias, but build a well filtered bias supply.

[Roscoe Primrose]

So don't regulate bias, but build a well filtered bias supply.

More accurately, don't regulate bias unless you regulate B+. Same holds true for screen voltage where applicable, it's an all-or-nothing situation...

Roscoe

[SoundMods]

graph-6550eh-triode-anodecurves.jpg

I know this is understood by the experienced DIY'ers, but was trying to explain to a relative NOOB why using a LM417HV in a bias circuit was a bad idea. The attached figure I drew up illustrates the problem. Please ignore the bad operating point I chose (above the plate dissipation limit), it was purely for purposes of discussion.

In this particular example, an increase in line voltage from 117V to 125V (a 7% change), produces a 48% increase in plate current with a regulated fixed bias, whereas with a filtered bias supply (floating with voltage changes), there is an 18% increase in plate current. Obviously, you would not expect to see a 7% variance in voltage, but is not unusual to see fluctuations of 2 or 3 volts depending on the time of day and load. Regulated bias will result is a pretty significant change in operating characteristics.

So don't regulate bias, but build a well filtered bias supply.

Or as Sid Smith (Radio Craftsman & Marantz) postulated cathode bias get's the job done and tracks the audio signal. He actually preferred one cathode bias resistor for a pair of output tubes. And of course those that know me know that you cannot skimp on cathode bypass caps. High quality film as opposed to electrolytics.

[dberning]

A triode-connected 6550 is not as low in impedance as tubes such as the 300B, and the situation becomes worse with low-impedance triodes. Using fixed bias for these tubes is risky, and better to use a combination of cathode bias and fixed bias for a simple solution. Best to use servo fixed bias if maximum performance is to be had, but properly done circuits are tricky. Sample and hold techniques in the analog domain or digital with storage are both methods I have used.

If fixed bias is used, it is possible with a simple bias network to cancel the affect of the line voltage variations changing the plate current. For example, a series zener diode making up part of the bias network would make the bias voltage have a larger percentage change than the plate voltage, whereas no zener would result in the plate voltage changing with the same percentage as the bias voltage. But with low-impedance triodes, it is good practice to have some amount of cathode degeneration for stability unless a good servo bias is used.

David

[SoundMods]

A triode-connected 6550 is not as low in impedance as tubes such as the 300B, and the situation becomes worse with low-impedance triodes. Using fixed bias for these tubes is risky, and better to use a combination of cathode bias and fixed bias for a simple solution. Best to use servo fixed bias if maximum performance is to be had, but properly done circuits are tricky. Sample and hold techniques in the analog domain or digital with storage are both methods I have used.

If fixed bias is used, it is possible with a simple bias network to cancel the affect of the line voltage variations changing the plate current. For example, a series zener diode making up part of the bias network would make the bias voltage have a larger percentage change than the plate voltage, whereas no zener would result in the plate voltage changing with the same percentage as the bias voltage. But with low-impedance triodes, it is good practice to have some amount of cathode degeneration for stability unless a good servo bias is used.

David

In the good ole days G.E. used that combination of fixed and cathode bias. More recently Jadis used that scheme in their Defy-7 with a brace of (6) 6550s per channel. The idea is to have a starting fixed-bias setting with the cathode bias working with the signal.