OP
-
Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
Phono Cartridge Response Measurement Script
- Thread starter JP
- Start date
OP
- Thread Starter
- #1,523
Did you migrate or start fresh? If the former, likely easiest to wipe out your Python install and start over.
OP
- Thread Starter
- #1,524
GitHub - FidelisAnalog/cartridge-loading-calculator
Contribute to FidelisAnalog/cartridge-loading-calculator development by creating an account on GitHub.
github.com
First stab at electrical loading "calculator".
Last edited:
- Joined
- Dec 26, 2017
- Messages
- 2,700
- Likes
- 2,348
OP
- Thread Starter
- #1,527
Moving this here.
I had Claude compare the Python divider model to this ^ JS implementation. The punchline is that the LLM thinks the JS implementation is calculating impedance (admittance) so a current source model. I haven't verified this.
OP
- Thread Starter
- #1,528
Thanks but I need start of left sweep to start of right pilot, not start of pilot to start of pilot.
ElectricWarrior
Member
OP
- Thread Starter
- #1,530
Regardless, the entire exercise reminds me of why I've always discounted the approach, at least with the simple model that can be done with the available data. Real response for various Cl:
Model:
It's difficult to align the influence of the electrical parameters of the load on to real measurements.
GuidoK
Member
It could be interesting if you could put in measured inductance at 1khz and 10khz (most cheap lcr meters can measure those frequencies nowadays).
The 2 coil AT cartridges like the at-vm95 series have a significantly lower inductance at 10khz (probably due to core saturation) and therefore at 10khz the efficiency is less (which means less output).
I expect that the laminated cores in the at-vm5xx and vm7xx series (and for instance in shure's v15 series) is to prevent that (they have much less drop in inductance at 10khz)
OP
- Thread Starter
- #1,532
Load capacitance on MM cartridge
I'm going to build my own RIAA pre-amp. It will be fitted inside the turntable. My MM cartridge (Ortofon 2M Red) specifies the load capacitance as 150 - 300 pF. Am I better of aiming for the low side (150 pF), the high side (300 pF) or something in the middle? Besides the tone-arm's cable...
I was exploring the six component model and split inductance when my AI credits ran out. Interesting conversation with a fake, mostly smart, drunk uncle. Stray winding capacitance didn't pan out. I also need to look again at Hans work. I know I had issues with it but don't recall the specifics.
At the end of the day I don't know if it's worthwhile. The loading changes tend to have a linear enough impact on the response that you don't need much more than two point of C and/or R along with knowledge of where the corner of the LPF sits for a reasonable prediction of optimal load. On most carts Cl is about 0.75-1dB/100pF.
OP
- Thread Starter
- #1,533
SJPlot current version now 18.6.4. Featured changes:
As a performance benchmark I ran version 18.3.8 standalone on my desktop for TRS1007, sweep extraction, and bass emphasis applied. Total processing time was 6.94 seconds. The same configuration for 18.6.5 was 1.96 seconds. Not bad.
- plot_data_out, if enabled, now outputs to CSV file - one per channel. The file(s) are named after plot_info in the same way that the plot image file is, and are placed in the directory the script is run from.
- Normalize indicator will only be placed on the plot if it falls within the plot data range. This is for when start_f and end_f are used so the plot will scale properly to the data range. Along with this change, on the web UI there is no longer a need to ensure the normalize value is between start_f and end_f, so this logic has been removed.
- Plot data download button added to web UI.
As a performance benchmark I ran version 18.3.8 standalone on my desktop for TRS1007, sweep extraction, and bass emphasis applied. Total processing time was 6.94 seconds. The same configuration for 18.6.5 was 1.96 seconds. Not bad.
| Component | Old | New | Speedup | Time Saved |
|---|---|---|---|---|
| Sweep extraction | 3.33s | 0.43s | 7.7× | -2.90s (58%) |
| createplotdata (left) | 1.33s | 0.27s | 4.9× | -1.06s (21%) |
| createplotdata (right) | 1.25s | 0.26s | 4.8× | -0.99s (20%) |
| Matplotlib | 0.69s | 0.66s | 1.04× | -0.03s (1%) |
| TOTAL | 6.94s | 1.96s | 3.5× | -4.98s (72%) |
Last edited:
Total newb here. Is there a step-by-step perhaps video or guide for new users? I have the Feickert Adjust+ and am thinking that perhaps I am not getting completely accurate frequency response curves and want to verify with this method. Can I simply record the sweep into Audacity and input into the SJPlot without downloading any script?
OP
- Thread Starter
- #1,535
SJPlot Online
If you want automatic processing you'll need to use one of the listed test records, preferably (CA)-TRS-1007 or STR 100 being the most accurate. To use a sweep from the Feickert record you'll need to pre-process the audio and use the tool in Advanced mode. Instructions for the fields are on GitHub (link at bottom of the page). I don't have the record so I can't tell you what the specific steps would be.
OP
- Thread Starter
- #1,536
Was searching the thread for something else and came across this post, which I'd forgotten. If you or anyone else wants to push this please feel free. At this point it is feature complete - the only thing it doesn't do that the standalone script does is save the extracted sweeps to files.
Thank you! So, is automatic processing compatible with a Mac?SJPlot Online
sjplot.com
If you want automatic processing you'll need to use one of the listed test records, preferably (CA)-TRS-1007 or STR 100 being the most accurate. To use a sweep from the Feickert record you'll need to pre-process the audio and use the tool in Advanced mode. Instructions for the fields are on GitHub (link at bottom of the page). I don't have the record so I can't tell you what the specific steps would be.
OP
- Thread Starter
- #1,538
I update the app location, requirements, and resources in post #2, and created sjplot.com/online basic troubleshooting and help request steps in post #4.
If you want to get in to the dirt on how everything works, Claude and I tore it up last night.
If you want to get in to the dirt on how everything works, Claude and I tore it up last night.
OP
- Thread Starter
- #1,539
I've completed the initial work in moving a lot of the SJPlot components in to a custom library so that they're more easily reusable and so that maintenance/enhancements are abstracted to a single place. The initial goal of this restructuring is to publish tooling that facilitates analysis without the strict one or two channel FR plot use case, such as multi-dimensional comparisons.
This also means the core functions can easily be used for other things; you can pass audio to the library function that analyzes the sweeps and get the data back, and then do whatever you want with it.
I'm sticking to the signal pipeline tenants of signal conditioning -> data creation -> data conditioning, however the order of operations within each category is subject to be changed, or is going to be changed, and in at least one known case these changes will alter the data being plotted. As this is bound to a data smoothing filter operation that has always been used in the script, the degree of impact is not new due to the nature of data smoothing - neighboring data will always influence each other. This also means the absolute change cannot be predicted as it is dependent on the input data. Also not a new thing.
The only reason I'm raising it is for awareness that data from this new architecture and the old isn't going to perfectly align, but the misalignment is benign in that the only thing that's really happening is you being able to compare a function doing what it has always done.
More to come.
This also means the core functions can easily be used for other things; you can pass audio to the library function that analyzes the sweeps and get the data back, and then do whatever you want with it.
I'm sticking to the signal pipeline tenants of signal conditioning -> data creation -> data conditioning, however the order of operations within each category is subject to be changed, or is going to be changed, and in at least one known case these changes will alter the data being plotted. As this is bound to a data smoothing filter operation that has always been used in the script, the degree of impact is not new due to the nature of data smoothing - neighboring data will always influence each other. This also means the absolute change cannot be predicted as it is dependent on the input data. Also not a new thing.
The only reason I'm raising it is for awareness that data from this new architecture and the old isn't going to perfectly align, but the misalignment is benign in that the only thing that's really happening is you being able to compare a function doing what it has always done.
More to come.
Last edited:
Similar threads
