I've been meddling around with a TDA1541 analogmetric dac. The infamous one. And I've been pondering the idea of letting the -15v/+15v opamp voltages float according to line voltage. Simply put, I've removed the typical LM317/337 voltage regulators and in place I've placed a 47uH/0.1ohm choke and a 100 ohm resistor wired in series with the choke. The power then goes directly into a 100uF capacitor (Elna Cerafine) to add as an LC/RC hybrid choke and shunt noise to ground. The 100 ohm resistor in series with the choke is used to drop the voltage from 17.5v down to 15v.
I've been regularly monitoring the voltage fluctuations according to line voltage and everything is within the tolerances of the AD797 op amps that I am using. The maximum voltage of 18v of the ad797 makes this modification novel.
I was wondering if anyone else has done this before besides me? If so what benefits were there. I figured that without the voltage regulator in the circuit the noise levels would go to the uV range and without the voltage regulator in place there is no produced noise aside from thermal noise generated from the resistors themselves. Any line noise can then be filtered out by a mains filter socket (I have a NEC/TOKIN GL-2080FVP-T1) or by the filtering caps and the 100nF ceramic capacitors nearby the opamps.
The circuit is as follows:
2x 12v AC windings on a dedicated R-Core transformer.
2x Germanium Diodes (0.3v vdrop)
+
1x Panasonic FC 3900uF capacitor (I found that these sounded best, compared to Elna Cerafines.)
+
47uH choke (measures 0.1ohm) + 100 ohm 1/4 watt metal film resistor in series.
+
100uF Elna Cerafine decoupling capacitor.
+
2x per rail 100nF Ceramic glass diffused decoupling cap near to the AD797 op amps.
+
-15v or +15v Op amps input.
The line voltage will swing from 249v down to 240v during the morning/day and as far as I can tell the only fluctuations on the output side of the filtering circuit is a voltage fluctuation from 14.80v to 15.20v. For example the line voltage is currnetly 244.9v and the output is 15.27v.
I've also considered doing this to my TDA1541 chip for both 5v and 15v rails but have been hesitant to do so because of the maximum 15v voltage of 16v. And because I think that the fluctuating voltages would cause havoc with the digital stages. It is a bit risky and requires constant monitoring but if you have the DAC powered by a UPS the line voltage cannot effect the final voltage too much because the UPS cuts in if the voltage gets too far out of range. Also my line voltage is very stable as it is.
I've been regularly monitoring the voltage fluctuations according to line voltage and everything is within the tolerances of the AD797 op amps that I am using. The maximum voltage of 18v of the ad797 makes this modification novel.
I was wondering if anyone else has done this before besides me? If so what benefits were there. I figured that without the voltage regulator in the circuit the noise levels would go to the uV range and without the voltage regulator in place there is no produced noise aside from thermal noise generated from the resistors themselves. Any line noise can then be filtered out by a mains filter socket (I have a NEC/TOKIN GL-2080FVP-T1) or by the filtering caps and the 100nF ceramic capacitors nearby the opamps.
The circuit is as follows:
2x 12v AC windings on a dedicated R-Core transformer.
2x Germanium Diodes (0.3v vdrop)
+
1x Panasonic FC 3900uF capacitor (I found that these sounded best, compared to Elna Cerafines.)
+
47uH choke (measures 0.1ohm) + 100 ohm 1/4 watt metal film resistor in series.
+
100uF Elna Cerafine decoupling capacitor.
+
2x per rail 100nF Ceramic glass diffused decoupling cap near to the AD797 op amps.
+
-15v or +15v Op amps input.
The line voltage will swing from 249v down to 240v during the morning/day and as far as I can tell the only fluctuations on the output side of the filtering circuit is a voltage fluctuation from 14.80v to 15.20v. For example the line voltage is currnetly 244.9v and the output is 15.27v.
I've also considered doing this to my TDA1541 chip for both 5v and 15v rails but have been hesitant to do so because of the maximum 15v voltage of 16v. And because I think that the fluctuating voltages would cause havoc with the digital stages. It is a bit risky and requires constant monitoring but if you have the DAC powered by a UPS the line voltage cannot effect the final voltage too much because the UPS cuts in if the voltage gets too far out of range. Also my line voltage is very stable as it is.
Last edited:
The primary benefit of an LM317 type regulator is accurate D.C. supply voltage and rejection of input ripple pass through. Better regulators are also effective at rejecting higher frequency supply line noise. High frequency output impedance (as seen by the load) is largely determined by the regulator's discrete shunt output capacitor.
Some circuits require an accurate D.C. supply voltage, such as logic chips. Other chips not so much, such as op-amp chips and other linear circuits. Many circuits are sensitive to supply noise, depending on their inherent power supply rejection ratio (PSRR).
Whether the regulator can be omitted depends on the specific factors mentioned above. Is an accurate D.C. voltage required? Is there significant noise or hash present on the supply line input which needs rejecting? Does the load itself induce significant fluctuation of the supply? Does the load feature an high PSRR across the band of interest and is therefore relatively insensitive to supply noise? Etc. Passive LCR supply filtering can be very effective as input noise frequency rises, but typically offer litter help at ripple frequencies and none at D.C.
Some circuits require an accurate D.C. supply voltage, such as logic chips. Other chips not so much, such as op-amp chips and other linear circuits. Many circuits are sensitive to supply noise, depending on their inherent power supply rejection ratio (PSRR).
Whether the regulator can be omitted depends on the specific factors mentioned above. Is an accurate D.C. voltage required? Is there significant noise or hash present on the supply line input which needs rejecting? Does the load itself induce significant fluctuation of the supply? Does the load feature an high PSRR across the band of interest and is therefore relatively insensitive to supply noise? Etc. Passive LCR supply filtering can be very effective as input noise frequency rises, but typically offer litter help at ripple frequencies and none at D.C.
The HF noise reduces with larger regulator output capacitance. Also, you can use
an RC filter after the regulator.
Last edited:
Everything you've done to "improve noise levels" applies only to high frequency noise. If you want to also "improve noise levels" at low frequencies, install a voltage regulator IC between the rectifiers and your passive filters. The voltage regulator IC completely eliminates low frequency noise, and your passive filters do their best at high frequencies.
Yes I know all of that. I've been skimming over a few websites detailing opamps and resistor noise. Specifically this one: Working with Cranky Op-Amps
What people here have missed though is that the opamp itself has its own PSRR. So there is no need for a voltage regulator for the removal of noise from the PSU purely from the PSRR standpoint. PROVIDED that you get the ripple low before you feed the opamp. I hope to get it low enough so it is inaudible.
The opamp itself removes the noise but only to a certian extent. I believe that with the 100nF capacitors in place it will provide a low enough 50hz/100hz rejection so that its inaudible. But I have yet to complete the dac to find out.
It would be interesting for anyone else out there who has done this to their opamp stages and discovered beneficial results. I haven't yet completed my DAC, this is my 2nd revision of the TDA1541. I've previously built TDA1543 dacs and loved the sound when they were fed by an E-core transformer followed by a vreg and a supercapacitor on the vregs output limited by a resistor (so it doesn't blow the vreg), and a resistor to bleed the supercap once power is removed. I compared a R-Core with an E-core and preferred the E-core sonically.
What people here have missed though is that the opamp itself has its own PSRR. So there is no need for a voltage regulator for the removal of noise from the PSU purely from the PSRR standpoint. PROVIDED that you get the ripple low before you feed the opamp. I hope to get it low enough so it is inaudible.
The opamp itself removes the noise but only to a certian extent. I believe that with the 100nF capacitors in place it will provide a low enough 50hz/100hz rejection so that its inaudible. But I have yet to complete the dac to find out.
It would be interesting for anyone else out there who has done this to their opamp stages and discovered beneficial results. I haven't yet completed my DAC, this is my 2nd revision of the TDA1541. I've previously built TDA1543 dacs and loved the sound when they were fed by an E-core transformer followed by a vreg and a supercapacitor on the vregs output limited by a resistor (so it doesn't blow the vreg), and a resistor to bleed the supercap once power is removed. I compared a R-Core with an E-core and preferred the E-core sonically.
Last edited:
You can add a resistor in parallel with the psu output to keep voltages under check, its inefficient but it works. I can adjust the voltage down from 15.20v to 14.85v with just a 2k2 1/4 watt.
100nF will not do anything with 50/100Hz. Did you mean 100uF? 100uF plus 100 ohms (the 47uH is irrelevant) will attenuate 100Hz by about 16dB.VenusFly said:
It is quite possible that your new setup has more audio frequency noise than the old one. It may have less high frequency noise, but you can't hear this anyway.
You will need to explain what you mean by this, with a diagram.
You may now be approaching the level of ripple that the regulator would have given you.
Be aware that ripple as calculated by PSUD2 does not include the effects of grounding, which can dominate at smaller voltages. A very common newbie error is to use big chokes and big caps to get infinitesimal levels of ripple (in simulation) and then find that in real life there is hum/buzz because of poor grounding.
Also, a 3H choke will not stop high frequency interference as at higher frequencies it looks like a capacitor.
I am sure you will eventually learn a lot from this exercise, but it is unlikely to improve the sound quality.
Be aware that ripple as calculated by PSUD2 does not include the effects of grounding, which can dominate at smaller voltages. A very common newbie error is to use big chokes and big caps to get infinitesimal levels of ripple (in simulation) and then find that in real life there is hum/buzz because of poor grounding.
Also, a 3H choke will not stop high frequency interference as at higher frequencies it looks like a capacitor.
I am sure you will eventually learn a lot from this exercise, but it is unlikely to improve the sound quality.
Do you mean something like the Aikido circuit? The power ripple AC is fed through a capacitor to the tube stage in opposite phase, so that it cancels hum at the output.
I didn't have a name for it I just called it a humbucker circuit.
Now that I do I'll look it up! Thanks.
I just powered up the dac and I must say that this analogmetric pcb is complete garbage. It wasn't difficult with a small pencil iron to burn and melt the solder tabs off and around the holes, just by the mere act of heating them.
First CD was shakespeares sisters and it sounded glorious. However it did have some crackling going on, which is probably because I haven't yet broken the trace which gives the secondary AD797s a feedback loop. I also haven't put the 2200pF capacitors in the PCB as detailed in the schematic so its running too hot.
The two final AD797 op amps were running really warm but the first two from the dac were running cool so I turned it off. I checked the outputs for short circuits but none were found and the voltages were fine, so its got to be the feedback loop which is overdriving the op amps.
The amazingly sounding vocals were the first thing that I noticed.
But think I'll plan on ditching this shitty low quality PCB and get myself something better.
First CD was shakespeares sisters and it sounded glorious. However it did have some crackling going on, which is probably because I haven't yet broken the trace which gives the secondary AD797s a feedback loop. I also haven't put the 2200pF capacitors in the PCB as detailed in the schematic so its running too hot.
The two final AD797 op amps were running really warm but the first two from the dac were running cool so I turned it off. I checked the outputs for short circuits but none were found and the voltages were fine, so its got to be the feedback loop which is overdriving the op amps.
The amazingly sounding vocals were the first thing that I noticed.
But think I'll plan on ditching this shitty low quality PCB and get myself something better.
yeah im totally giving up on this analogmetric dac. something is shorting out the opamps and making them warm-hot and draw a lot of current, I never did trust this pcb from the very start, you have to use a 5 watt pencil iron to not damage anything. I cannot find the issue and I'm not going to kill my AD797s. The current limiting resistor at 2k2 has saved them from damage.
I added the 2200pF caps and nothing changed. Theres a fault on the pcb or somewhere in my soldering, which is a nightmare with an unregulated soldering iron and this cheap nasty fiberglass/plastic pcb.
I'll go get myself a nice distinction-1541 v2 and call it a day. Gonna put my "diyaudio" hat up and give up for a while.
I added the 2200pF caps and nothing changed. Theres a fault on the pcb or somewhere in my soldering, which is a nightmare with an unregulated soldering iron and this cheap nasty fiberglass/plastic pcb.
I'll go get myself a nice distinction-1541 v2 and call it a day. Gonna put my "diyaudio" hat up and give up for a while.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.