Good to hear that you are (finally) making progress.
Not surprised by your diode result BTW. My recommendation was to update it to the Vbe multiplier solution but it would mean finding space somewhere for the bias pot and components.
If you do not get a good result with the 1N4448 the only other option I can suggest is BAV21.
That's got a 1V max at 100mA. 1S920 is 1.2V at 200mA.
1N4448 hopefully will work - spec is Vmin 0.62 Vmax 0.72 at 5mA.
Can't find any datasheets giving Vf:If for 1S920. Too old now.
Not that manudfacturers published useful graphs like a log plot. Only linear ones where you can't read the "small print".
Not surprised by your diode result BTW. My recommendation was to update it to the Vbe multiplier solution but it would mean finding space somewhere for the bias pot and components.
If you do not get a good result with the 1N4448 the only other option I can suggest is BAV21.
That's got a 1V max at 100mA. 1S920 is 1.2V at 200mA.
1N4448 hopefully will work - spec is Vmin 0.62 Vmax 0.72 at 5mA.
Can't find any datasheets giving Vf:If for 1S920. Too old now.
Not that manudfacturers published useful graphs like a log plot. Only linear ones where you can't read the "small print".
Well this is interesting......
Built the R board up last night with exactly the same components as the L board, unfortunately our friends at Farnell seem to have lost my 1N4448s in the post - so I though 'what the hell' I'll put 1N4148s in instead.
Amplifier works perfectly!!!
I'm going to check (later) the forward voltage of the 1N4148s I used in both boards to see if they differ substantially - they came from the same tape so I would expect them to be within certain parameters.
Built the R board up last night with exactly the same components as the L board, unfortunately our friends at Farnell seem to have lost my 1N4448s in the post - so I though 'what the hell' I'll put 1N4148s in instead.
Amplifier works perfectly!!!
I'm going to check (later) the forward voltage of the 1N4148s I used in both boards to see if they differ substantially - they came from the same tape so I would expect them to be within certain parameters.
It may be obvious in hindsight but your 1N4148's may not have a high enough forward voltage. ejp didn't explain why 4148's didn't work.
Simulations show that the base to base voltage of the first driver stage is about 1.39V. That needs two diodes at 0.695V which has to be at 3mA. That is around the top end of the 1N4448 spec. (0.72V at 5mA). So I may have been wrong to have suggested 4448's. Sorry! (And if I were a manufacturer I'd probably be making 4148's and 4448's on the same line, suggests they may not be very different, but 4148's will have 4448 rejects)
Trouble is exact Vbe matching is a challenge at the best of times since manufacturing lines and manufacturers' processes are movable objects. So the Vbe's of the drivers are an uncertain quantity. Then that is added to because of the 100 ohm emitter resistors and 1k base resistors on the next pair of transistors, adding maybe 180mV to the proverbial 1.2V, which is close to the simulated result.
If need be you can increase the bias voltage with a small resistor (physically and electrically). The bias current range with a 22k pot in the original 303 is about 1-3.5mA which is about 30mV per diode (or maybe up to 60 depending on the gold doping used. Last time I measured a 4148 the (non)ideality factor was about 1.5) so to bump up the voltage in 60mV jumps needs resistors of about 56 ohms per step. I'd try smaller resistors (22 ohms say) as I'd want the VAS current to be closer to 3mA than 1. (This resistor is in series with the diodes, obvs.)
You can measure the base-to-base voltages between the two boards to check this, I think.
Simulations show that the base to base voltage of the first driver stage is about 1.39V. That needs two diodes at 0.695V which has to be at 3mA. That is around the top end of the 1N4448 spec. (0.72V at 5mA). So I may have been wrong to have suggested 4448's. Sorry! (And if I were a manufacturer I'd probably be making 4148's and 4448's on the same line, suggests they may not be very different, but 4148's will have 4448 rejects)
Trouble is exact Vbe matching is a challenge at the best of times since manufacturing lines and manufacturers' processes are movable objects. So the Vbe's of the drivers are an uncertain quantity. Then that is added to because of the 100 ohm emitter resistors and 1k base resistors on the next pair of transistors, adding maybe 180mV to the proverbial 1.2V, which is close to the simulated result.
If need be you can increase the bias voltage with a small resistor (physically and electrically). The bias current range with a 22k pot in the original 303 is about 1-3.5mA which is about 30mV per diode (or maybe up to 60 depending on the gold doping used. Last time I measured a 4148 the (non)ideality factor was about 1.5) so to bump up the voltage in 60mV jumps needs resistors of about 56 ohms per step. I'd try smaller resistors (22 ohms say) as I'd want the VAS current to be closer to 3mA than 1. (This resistor is in series with the diodes, obvs.)
You can measure the base-to-base voltages between the two boards to check this, I think.
Last edited:
Good to hear you are getting there but its all still a bit of an unknown because we talked about the critical bias voltage (developed by these diodes) as early as post #3 and I don't think we got anywhere 
I've no answer to that one I'm afraid. I would need to measure myself what was going on with both sets of different parts fitted.
Those diodes are run at about 3.5 milliamps and so you could if you wanted try running two in series from a 9 volt battery via a series 1k8 or 2k2 resistor and compare forward voltages.
Use one of each diode type in the series chain. That way you know the currents are identical and so any difference in voltage is down to the diode.
Those diodes are run at about 3.5 milliamps and so you could if you wanted try running two in series from a 9 volt battery via a series 1k8 or 2k2 resistor and compare forward voltages.
Use one of each diode type in the series chain. That way you know the currents are identical and so any difference in voltage is down to the diode.
It can really only be variations in Vf. Just measured a couple at 5mA and got 710mV and 707mV. Depending on the (non)ideality that may drop 10-15mV at 3mA. But this is in the 1N4448 spec. And the voltages seem marginal. Suspect you will need one of those resistors to push the volts up a bit.
Hi, I thought I would add my experience with similar issue. There are 2:
1) few years back I got a unit, it was the older style board with 4 diodes in a row. Could not set the bias. Eventually I bought new boards and components and all worked. But I still didnt like the fact I didnt find the cause
2) I have been fixing/upgrading a unit for a friend last and this week. Before doing any upgrades I teated transistors, plugged it in and could set all volatges except for bias on one channel. It would be close to 0mV, then jump to 60-90 mV with slightest touch. Tjis time I had the amp from point 1 to compare. I changed small transistors to bc. Electrolytics and Diodes as well. Still the same. What was left were resistors and small caps. Changed the caps from the new unit one by one. Obviously it was the last one. C109 was 68pF, where it should be 0.1 uF. Found 0.1 50V (there is max 35 volts in that section) and all perfect. Measured the 'good' board and it was out too, but not as much and still doing its function then (0.02 uF). Swapped it too obviously. I still have the old boards that I could not get working and so I measured those caps and they were the same as the bad board around 70pF. So it was the reason I could not set bias previously.
So check the C109 whem working on quad 303 to save yourself a lot of hassle
1) few years back I got a unit, it was the older style board with 4 diodes in a row. Could not set the bias. Eventually I bought new boards and components and all worked. But I still didnt like the fact I didnt find the cause
2) I have been fixing/upgrading a unit for a friend last and this week. Before doing any upgrades I teated transistors, plugged it in and could set all volatges except for bias on one channel. It would be close to 0mV, then jump to 60-90 mV with slightest touch. Tjis time I had the amp from point 1 to compare. I changed small transistors to bc. Electrolytics and Diodes as well. Still the same. What was left were resistors and small caps. Changed the caps from the new unit one by one. Obviously it was the last one. C109 was 68pF, where it should be 0.1 uF. Found 0.1 50V (there is max 35 volts in that section) and all perfect. Measured the 'good' board and it was out too, but not as much and still doing its function then (0.02 uF). Swapped it too obviously. I still have the old boards that I could not get working and so I measured those caps and they were the same as the bad board around 70pF. So it was the reason I could not set bias previously.
So check the C109 whem working on quad 303 to save yourself a lot of hassle
Sorry, wrote it in bed without the schematics and my memory was misleading me. Yes it was c108. I don't know the theory behind, but when the original which measured
68pF was in, I could not set the bias, with a correct value cap in I could. And the cap was out of value on all 4 boards I worked on (one of them not critically)