For me the optima were just before 12 o'clock and around 2 o'clock, but they can vary depending on the matching between the triodes, so it's not surprising that you get different positions. It is a pity that your RV2 is completely clockwise, though, as that shows that the trimming range is too narrow.
I should have mentioned R14, R25, R26, R27, RV1, R103, R104, R105, R106 and RV2 in the section about things that may need to be changed experimentally, because I couldn't predict the required trimming range. Determining it experimentally would have involved buying a few dozens of E88CCs from many different brands, determining the optimal settings for those and doing some statistical calculations, but I didn't fancy buying dozens of E88CCs, so I ended up simply dimensioning things such that the trimming range was considerably larger than was needed in my DAC. I'll come up with a set of alternative values that double the trimming range.
The original valve DAC has no RV3 and RV4, but it used to have trimmer capacitors that had a similar influence on the timing as RV3 and RV4. As far as I recall, the exact setting did not have much influence on the noise floor until it was so far off that the DAC did not work as a return-to-zero DAC anymore, then the noise increased clearly.
Attachments
Thanks as always Marcel. I've been thinking (generally a dangerous activity)...
I was wondering about pulling all the tubes mixing them up and seeing if I can get a different result, (or not have a potentiometer fully rotated)? I guess it'll be pot luck but you never know.
Should the value of RV1/2 be amended for future builds? The downside is that you potentially make finding the 'sweet spot' even harder to nail with single turn pots because the adjustments will need to be much finer. Belay that - I've just reread your post and see you're already on the case.
Out of interest, did you get your DAC to have effectively zero noise or do you expect there to be a residual level that you can't drop below?
Your comments about RV3/4 align with my practical findings this morning when I couldn't hear any difference when I backed them off from full-clockwise. I don't want to get to the point where it suddently stops functioning properly and makes the loud noise. If adjusting RV3/4 is tending towards a binary scenario is there a downside to just leaving them full-clockwise?
I was wondering about pulling all the tubes mixing them up and seeing if I can get a different result, (or not have a potentiometer fully rotated)? I guess it'll be pot luck but you never know.
Should the value of RV1/2 be amended for future builds? The downside is that you potentially make finding the 'sweet spot' even harder to nail with single turn pots because the adjustments will need to be much finer. Belay that - I've just reread your post and see you're already on the case.
Out of interest, did you get your DAC to have effectively zero noise or do you expect there to be a residual level that you can't drop below?
Your comments about RV3/4 align with my practical findings this morning when I couldn't hear any difference when I backed them off from full-clockwise. I don't want to get to the point where it suddently stops functioning properly and makes the loud noise. If adjusting RV3/4 is tending towards a binary scenario is there a downside to just leaving them full-clockwise?
Last edited:
Mixing up the valves could indeed help.
My DAC has noise floors of -85.76 dB(A) on the left and -91.29 dB(A) on the right channel when I use the sigma-delta modulator in the PWM8 setting, so with an embedded pulse-width modulator. It's a few dB worse in the other modes, PWM4 and chaotic single-bit. It's such a low noise level that I never heard it during normal use, but I can hear something when I turn up the volume knob a lot and put my ear right next to the tweeter.
Regarding RV3 and RV4, you could first set the volume low and check over what range you can vary RV3 and RV4 without getting any loud noises. When you did have a timing problem, there were other issues, basically the whole DAC was running at twice the intended frequency. The expectation is that RV3 and RV4 have to be set to the same position.
My DAC has noise floors of -85.76 dB(A) on the left and -91.29 dB(A) on the right channel when I use the sigma-delta modulator in the PWM8 setting, so with an embedded pulse-width modulator. It's a few dB worse in the other modes, PWM4 and chaotic single-bit. It's such a low noise level that I never heard it during normal use, but I can hear something when I turn up the volume knob a lot and put my ear right next to the tweeter.
Regarding RV3 and RV4, you could first set the volume low and check over what range you can vary RV3 and RV4 without getting any loud noises. When you did have a timing problem, there were other issues, basically the whole DAC was running at twice the intended frequency. The expectation is that RV3 and RV4 have to be set to the same position.
Good point. I think it would be a good idea if I made the PCB compatible with both, so people can choose single turn or multiturn.
There are multiturn trimmer potmeters that should fit into the PCB as is (Bourns 3296P series, Blocked ), but those have the adjustment screw on the side, facing the big DC blocking capacitors. I think those would be very inconvenient.
The change to double the resistance trimming range is to change R14, R25, R26, R27, R103, R104, R105 and R106 into 2 kohm +/- 1 %, 0.6 W metal film. The trimming potmeters can stay as is, as can R5, R16, R76 and R87.
There are multiturn trimmer potmeters that should fit into the PCB as is (Bourns 3296P series, Blocked ), but those have the adjustment screw on the side, facing the big DC blocking capacitors. I think those would be very inconvenient.
The change to double the resistance trimming range is to change R14, R25, R26, R27, R103, R104, R105 and R106 into 2 kohm +/- 1 %, 0.6 W metal film. The trimming potmeters can stay as is, as can R5, R16, R76 and R87.
Ray, could you try the AMSDM7 option in HQPlayer and see if this gets rid of the noise problem(s)?
There are actually three possibilities:
1. The DAC works as intended, that is, as a return to zero DAC where the upper differential pairs switch at moments when they have no tail current
2. The timing is so messed up that a flip-flop goes metastable or some of the data bits get sampled more often than others and you get very loud noise
3. The DAC works, but not as a return to zero DAC because there is too little delay between clock and clock5 / clockn and clockn5, so the data to the upper differential pairs already changes while they still have tail current
What you had in your first tests was apparently mode 2. The increased noise I wrote about in post 621 was due to the DAC going from mode 1 to mode 3. That gives a noticeable noise increase, but it is not even close to the noise of mode 2.
The purpose of the RV3/RV4 adjustment is to get the DAC to reliably work in mode 1 rather than mode 3. I expect that now that you have the correct jumper settings, you won't be able to get into mode 2 anymore, but it is good to check that before you put your head up against a loudspeaker.
If the trimming range of RV3 and RV4 should be large enough to get into mode 2, it could be that that only happens over a very small part of the range. So you have to turn the trimmer potmeters slowly from fully clockwise to fully anticlockwise to check if there is a region with loud noise.
Yes, the 5 kohm version of the PV37W series should fit without any PCB change. Thanks!
Will do though my understanding is that the noise (which I wouldn't describe as a problem as I already have it low enough not to be audible from my listening position) is down to small variances in the hardware elements of the project?
Those are rated at 0.25W whereas the orignal spec pots were 0.5W - I assume a non-issue?
Sorry, actually 0.25 W is a bit tight. If the DAC's data lines get disconnected and the potmeter is turned almost entirely to one side, there might be 10 mA flowing through a small part of the resistive element. The power spec of 0.25 W is for the entire resistive element, so the current the potmeter is designed for is sqrt(0.25 W/5 kohm) ~= 7.071068 mA which is less than 10 mA.
So, I'll have to update the PCB design after all.
So, I'll have to update the PCB design after all.
The valves and the data processing digital circuits are deliberately placed on different sides of the PCB, so the PCB ground and supply planes act as shields between them.
My build has the valves showing through the top plate - it's why I added the bling of LEDs under the valve sockets - I just moved the big components like electrolytics, output caps etc. to the underside of the board. The top plate for my chassis is being machined at the moment so I hope to be able to install it by next weekend.